From: Lei Xia Date: Mon, 15 Feb 2010 05:33:24 +0000 (-0600) Subject: Add lock to vmm_queue X-Git-Url: http://v3vee.org/palacios/gitweb/gitweb.cgi?p=palacios.git;a=commitdiff_plain;h=68521eca0f94f8c22aa60e762c83e24beec7233c Add lock to vmm_queue Update VNET and virtio NIC to support multiple guests --- diff --git a/palacios/include/palacios/vmm_queue.h b/palacios/include/palacios/vmm_queue.h index a4994f5..e88329f 100644 --- a/palacios/include/palacios/vmm_queue.h +++ b/palacios/include/palacios/vmm_queue.h @@ -25,7 +25,7 @@ #include #include - +#include /* IMPORTANT: @@ -43,9 +43,7 @@ struct queue_entry { struct gen_queue { uint_t num_entries; struct list_head entries; - - // We really need to implement this.... - // void * lock; + v3_lock_t lock; }; diff --git a/palacios/include/palacios/vmm_vnet.h b/palacios/include/palacios/vmm_vnet.h index 90e3ca1..eef724c 100644 --- a/palacios/include/palacios/vmm_vnet.h +++ b/palacios/include/palacios/vmm_vnet.h @@ -9,8 +9,6 @@ * * Copyright (c) 2009, Lei Xia * Copyright (c) 2009, Yuan Tang - * Copyright (c) 2009, Jack Lange - * Copyright (c) 2009, Peter Dinda * All rights reserved. * @@ -40,6 +38,8 @@ typedef enum {MAC_ANY, MAC_NOT, MAC_NONE} mac_type_t; //for 'src_mac_qual' and ' typedef enum {LINK_INTERFACE, LINK_EDGE, LINK_ANY} link_type_t; //for 'type' and 'src_type' in struct routing typedef enum {TCP_TYPE, UDP_TYPE, NONE_TYPE} prot_type_t; +#define VNET_INITAB_HCALL 0xca00 + //routing table entry struct routing_entry{ char src_mac[6]; @@ -115,7 +115,6 @@ struct vnet_if_link { }__attribute__((packed)); -//link table entry struct link_entry { link_type_t type; @@ -127,6 +126,13 @@ struct link_entry { int use; }__attribute__((packed)); +struct ethernet_pkt { + uint32_t size; //size of data field + uint16_t type; + uint8_t use_header; + struct udp_link_header ext_hdr; + char data[ETHERNET_PACKET_LEN]; +}__attribute__((packed)); int v3_vnet_send_rawpkt(uchar_t *buf, int len, void *private_data); int v3_vnet_send_udppkt(uchar_t *buf, int len, void *private_data); @@ -147,7 +153,7 @@ int vnet_add_route_entry(char src_mac[6], int v3_vnet_pkt_process(); -void v3_vnet_init(struct guest_info *vm); +void v3_init_vnet(); #endif diff --git a/palacios/src/devices/lnx_virtio_nic.c b/palacios/src/devices/lnx_virtio_nic.c index ea998af..e83a199 100644 --- a/palacios/src/devices/lnx_virtio_nic.c +++ b/palacios/src/devices/lnx_virtio_nic.c @@ -26,6 +26,7 @@ #include #include #include +#include #include @@ -37,28 +38,7 @@ //#define VIRTIO_NIC_PROFILE -/* The feature bitmap for virtio net */ -#define VIRTIO_NET_F_CSUM 0 /* Host handles pkts w/ partial csum */ -#define VIRTIO_NET_F_GUEST_CSUM 1 /* Guest handles pkts w/ partial csum */ -#define VIRTIO_NET_F_MAC 5 /* Host has given MAC address. */ -#define VIRTIO_NET_F_GSO 6 /* Host handles pkts w/ any GSO type */ -#define VIRTIO_NET_F_GUEST_TSO4 7 /* Guest can handle TSOv4 in. */ -#define VIRTIO_NET_F_GUEST_TSO6 8 /* Guest can handle TSOv6 in. */ -#define VIRTIO_NET_F_GUEST_ECN 9 /* Guest can handle TSO[6] w/ ECN in. */ -#define VIRTIO_NET_F_GUEST_UFO 10 /* Guest can handle UFO in. */ -#define VIRTIO_NET_F_HOST_TSO4 11 /* Host can handle TSOv4 in. */ -#define VIRTIO_NET_F_HOST_TSO6 12 /* Host can handle TSOv6 in. */ -#define VIRTIO_NET_F_HOST_ECN 13 /* Host can handle TSO[6] w/ ECN in. */ -#define VIRTIO_NET_F_HOST_UFO 14 /* Host can handle UFO in. */ -#define VIRTIO_NET_F_MRG_RXBUF 15 /* Host can merge receive buffers. */ -#define VIRTIO_NET_F_STATUS 16 /* virtio_net_config.status available */ -#define VIRTIO_NET_F_CTRL_VQ 17 /* Control channel available */ -#define VIRTIO_NET_F_CTRL_RX 18 /* Control channel RX mode support */ -#define VIRTIO_NET_F_CTRL_VLAN 19 /* Control channel VLAN filtering */ -#define VIRTIO_NET_F_CTRL_RX_EXTRA 20 /* Extra RX mode control support */ #define VIRTIO_NET_S_LINK_UP 1 /* Link is up */ - -/* Maximum packet size we can receive from tap device: header + 64k */ #define VIRTIO_NET_MAX_BUFSIZE (sizeof(struct virtio_net_hdr) + (64 << 10)) struct virtio_net_hdr { @@ -66,7 +46,6 @@ struct virtio_net_hdr { #define VIRTIO_NET_HDR_GSO_NONE 0 /* Not a GSO frame */ uint8_t gso_type; - uint16_t hdr_len; /* Ethernet + IP + tcp/udp hdrs */ uint16_t gso_size; /* Bytes to append to hdr_len per frame */ uint16_t csum_start; /* Position to start checksumming from */ @@ -94,6 +73,7 @@ struct virtio_net_state { struct virtio_net_config net_cfg; struct virtio_config virtio_cfg; + struct vm_device * dev; struct pci_device * pci_dev; int io_range_size; @@ -101,24 +81,24 @@ struct virtio_net_state { struct virtio_queue tx_vq; //index 1, svq in Linux virtio driver, handle packet from guest struct virtio_queue ctrl_vq; //index 2, ctrol info from guest + ulong_t pkt_sent, pkt_recv, pkt_drop; + struct gen_queue * inpkt_q; + struct v3_dev_net_ops * net_ops; void * backend_data; struct virtio_dev_state * virtio_dev; struct list_head dev_link; - - ulong_t pkt_sent, pkt_recv, pkt_drop; }; +struct eth_pkt { + uint32_t size; //size of data + char data[ETHERNET_PACKET_LEN]; +}__attribute__((packed)); -#ifdef CONFIG_DEBUG_VIRTIO_NET - -static void print_packet(uchar_t *pkt, int size) { - PrintDebug("Virtio Nic: print_data_packet: size: %d\n", size); - v3_hexdump(pkt, size, NULL, 0); -} - -#endif +//temporary store all of the virtio states +static struct virtio_net_state *temp_net_states[4]; +static int net_idx = 0; #if 0 //Temporarly for debug @@ -163,7 +143,7 @@ static int virtio_free(struct vm_device * dev) return -1; } -static int virtio_reset(struct virtio_net_state * virtio) +static int virtio_init_state(struct virtio_net_state * virtio) { virtio->rx_vq.ring_desc_addr = 0; virtio->rx_vq.ring_avail_addr = 0; @@ -187,6 +167,11 @@ static int virtio_reset(struct virtio_net_state * virtio) //virtio->virtio_cfg.status = VIRTIO_NET_S_LINK_UP; virtio->virtio_cfg.pci_isr = 0; + virtio->pkt_sent = virtio->pkt_recv = virtio->pkt_drop = 0; + + virtio->inpkt_q = v3_create_queue(); + v3_init_queue(virtio->inpkt_q); + return 0; } @@ -205,10 +190,6 @@ static int pkt_write(struct virtio_net_state * virtio, struct vring_desc * buf_d return -1; } - PrintDebug("Length=%d\n", buf_desc->length); - //PrintDebug("Buffer Descriptor (ptr=%p) gpa=%p, len=%d, flags=%x, next=%d, buf address: %p, send address: %p\n", buf_desc, - //(void *)(buf_desc->addr_gpa), buf_desc->length, buf_desc->flags, buf_desc->next, buf, virtio->net_ops->send); - if (virtio->net_ops->send(buf, len, (void *)virtio, NULL) == -1) { return -1; } @@ -240,7 +221,6 @@ static int copy_data_to_desc(struct virtio_net_state * virtio_state, struct vrin PrintError("Could not translate buffer address\n"); return -1; } - len = (desc->length < buf_len)?desc->length:buf_len; memcpy(desc_buf, buf, len); @@ -260,14 +240,12 @@ static int send_pkt_to_guest(struct virtio_net_state * virtio, uchar_t * buf, ui PrintDebug("VIRTIO NIC: sending packet to virtio nic %p, size:%d", virtio, size); virtio->pkt_recv ++; - if (!raw) { data_len -= hdr_len; } build_receive_header(&hdr, buf, 1); - //queue is not set yet if (q->ring_avail_addr == 0) { PrintError("Queue is not set\n"); return -1; @@ -290,7 +268,6 @@ static int send_pkt_to_guest(struct virtio_net_state * virtio, uchar_t * buf, ui } memcpy((void *)hdr_addr, &hdr, sizeof(struct virtio_net_hdr)); - if (offset >= data_len) { hdr_desc->flags &= ~VIRTIO_NEXT_FLAG; } @@ -299,20 +276,16 @@ static int send_pkt_to_guest(struct virtio_net_state * virtio, uchar_t * buf, ui struct vring_desc * buf_desc = &(q->desc[buf_idx]); uint32_t len = 0; - len = copy_data_to_desc(virtio, buf_desc, buf + offset, data_len - offset); - + len = copy_data_to_desc(virtio, buf_desc, buf + offset, data_len - offset); offset += len; - if (offset < data_len) { buf_desc->flags = VIRTIO_NEXT_FLAG; } - - buf_desc->length = len; // TODO: do we need this? + buf_desc->length = len; } q->used->ring[q->used->index % q->queue_size].id = q->avail->ring[q->cur_avail_idx % q->queue_size]; q->used->ring[q->used->index % q->queue_size].length = data_len + hdr_len; // This should be the total length of data sent to guest (header+pkt_data) - q->used->index++; int last_idx = q->cur_avail_idx; @@ -321,17 +294,12 @@ static int send_pkt_to_guest(struct virtio_net_state * virtio, uchar_t * buf, ui q->idx_overflow = false; } else { virtio->pkt_drop++; - -#ifdef VIRTIO_NIC_PROFILE - PrintError("Virtio NIC: %p, one pkt dropped receieved: %ld, dropped: %ld, sent: %ld curidx: %d, avaiIdx: %d\n", virtio, virtio->pkt_recv, virtio->pkt_drop, virtio->pkt_sent, q->cur_avail_idx, q->avail->index); -#endif - } #ifdef VIRTIO_NIC_PROFILE - if ((virtio->pkt_recv % 10000) == 0){ - PrintError("Virtio NIC: %p, receieved: %ld, dropped: %ld, sent: %ld curidx: %d, avaiIdx: %d\n", virtio, virtio->pkt_recv, virtio->pkt_drop, virtio->pkt_sent, q->cur_avail_idx, q->avail->index); - } + PrintError("Virtio NIC: %p, one pkt dropped receieved: %ld, dropped: %ld, sent: %ld curidx: %d, avaiIdx: %d\n", + virtio, virtio->pkt_recv, virtio->pkt_drop, virtio->pkt_sent, q->cur_avail_idx, q->avail->index); #endif + } if (!(q->avail->flags & VIRTIO_NO_IRQ_FLAG)) { PrintDebug("Raising IRQ %d\n", virtio->pci_dev->config_header.intr_line); @@ -339,6 +307,14 @@ static int send_pkt_to_guest(struct virtio_net_state * virtio, uchar_t * buf, ui virtio->virtio_cfg.pci_isr = 0x1; } +#ifdef VIRTIO_NIC_PROFILE + if ((virtio->pkt_recv % 10000) == 0){ + PrintError("Virtio NIC: %p, receieved: %ld, dropped: %ld, sent: %ld\n", + virtio, virtio->pkt_recv, virtio->pkt_drop, virtio->pkt_sent); + } +#endif + + return offset; } @@ -348,19 +324,74 @@ int virtio_send(struct guest_info * vm, char *dev_name, uchar_t * buf, uint_t si struct vm_device * virtio_dev = v3_find_dev(vm, dev_name); // TODO: how to get virtio net state from device?? - // this is not right now struct virtio_net_state * virtio_state = (struct virtio_net_state *)virtio_dev->private_data; return send_pkt_to_guest(virtio_state, buf, size, 1, NULL); } +int __virtio_sendto_buf(struct virtio_net_state * net_state, uchar_t * buf, uint_t size) { + struct eth_pkt *pkt; + + pkt = (struct eth_pkt *)V3_Malloc(sizeof(struct eth_pkt)); + if(pkt == NULL){ + PrintError("Virtio NIC: Memory allocate fails\n"); + return -1; + } + + pkt->size = size; + memcpy(pkt->data, buf, size); + v3_enqueue(net_state->inpkt_q, (addr_t)pkt); + + PrintDebug("Virtio NIC: __virtio_sendto_buf: transmitting packet: (size:%d)\n", (int)pkt->size); + + return pkt->size; +} + + static int __virtio_dev_send(uchar_t * buf, uint32_t size, void *private_data) { struct virtio_net_state *virtio_state = (struct virtio_net_state *)private_data; - return send_pkt_to_guest(virtio_state, buf, size, 1, NULL); + return __virtio_sendto_buf(virtio_state, buf, size); +} + +/* + *called in svm/vmx handler + *iteative handled the unsent packet in incoming packet queues for + *all virtio nic devices in this guest + */ +int v3_virtionic_pktprocess(struct guest_info * info) +{ + struct eth_pkt *pkt = NULL; + struct virtio_net_state *net_state; + int i; + + for (i = 0; i < net_idx; i++) { + while (1) { + net_state = temp_net_states[i]; + if(net_state->dev->vm != info) + break; + + pkt = (struct eth_pkt *)v3_dequeue(net_state->inpkt_q); + if(pkt == NULL) + break; + + if (send_pkt_to_guest(net_state, pkt->data, pkt->size, 1, NULL)) { + PrintDebug("Virtio NIC: %p In pkt_handle: send one packet! pt length %d\n", + net_state, (int)pkt->size); + } else { + PrintDebug("Virtio NIC: %p In pkt_handle: Fail to send one packet, pt length %d, discard it!\n", + net_state, (int)pkt->size); + } + + V3_Free(pkt); + } + } + + return 0; } + static int get_desc_count(struct virtio_queue * q, int index) { struct vring_desc * tmp_desc = &(q->desc[index]); int cnt = 1; @@ -384,18 +415,12 @@ static int handle_pkt_tx(struct virtio_net_state * virtio_state) struct virtio_queue * q = &(virtio_state->tx_vq); struct virtio_net_hdr * hdr = NULL; - virtio_state->pkt_sent ++; - if (q->avail->index < q->last_avail_idx) q->idx_overflow = true; q->last_avail_idx = q->avail->index; -#ifdef VIRTIO_NIC_PROFILE - if(virtio_state->pkt_sent % 10000 == 0) - PrintError("Virtio NIC: %p, pkt_sent: %ld curidx: %d, avaiIdx: %d\n", virtio_state, virtio_state->pkt_sent, q->cur_avail_idx, q->avail->index); -#endif - - while (q->cur_avail_idx < q->avail->index ||(q->idx_overflow && q->cur_avail_idx < (q->avail->index + 65536))) { + while (q->cur_avail_idx < q->avail->index || + (q->idx_overflow && q->cur_avail_idx < (q->avail->index + 65536))) { struct vring_desc * hdr_desc = NULL; addr_t hdr_addr = 0; uint16_t desc_idx = q->avail->ring[q->cur_avail_idx % q->queue_size]; @@ -422,10 +447,10 @@ static int handle_pkt_tx(struct virtio_net_state * virtio_state) req_len += buf_desc->length; desc_idx = buf_desc->next; } + virtio_state->pkt_sent ++; q->used->ring[q->used->index % q->queue_size].id = q->avail->ring[q->cur_avail_idx % q->queue_size]; q->used->ring[q->used->index % q->queue_size].length = req_len; // What do we set this to???? - q->used->index++; int last_idx = q->cur_avail_idx; @@ -439,10 +464,14 @@ static int handle_pkt_tx(struct virtio_net_state * virtio_state) virtio_state->virtio_cfg.pci_isr = 0x1; } +#ifdef VIRTIO_NIC_PROFILE + if(virtio_state->pkt_sent % 10000 == 0) + PrintError("Virtio NIC: %p, pkt_sent: %ld\n", virtio_state, virtio_state->pkt_sent); +#endif + return 0; } - static int virtio_setup_queue(struct virtio_net_state * virtio_state, struct virtio_queue * queue, addr_t pfn, addr_t page_addr) { queue->pfn = pfn; @@ -454,19 +483,16 @@ static int virtio_setup_queue(struct virtio_net_state * virtio_state, struct vir // round up to next page boundary. queue->ring_used_addr = (queue->ring_used_addr + 0xfff) & ~0xfff; - if (guest_pa_to_host_va(virtio_state->virtio_dev->vm, queue->ring_desc_addr, (addr_t *)&(queue->desc)) == -1) { PrintError("Could not translate ring descriptor address\n"); return -1; } - if (guest_pa_to_host_va(virtio_state->virtio_dev->vm, queue->ring_avail_addr, (addr_t *)&(queue->avail)) == -1) { PrintError("Could not translate ring available address\n"); return -1; } - if (guest_pa_to_host_va(virtio_state->virtio_dev->vm, queue->ring_used_addr, (addr_t *)&(queue->used)) == -1) { PrintError("Could not translate ring used address\n"); return -1; @@ -483,41 +509,33 @@ static int virtio_setup_queue(struct virtio_net_state * virtio_state, struct vir return 0; } - - static int virtio_io_write(uint16_t port, void * src, uint_t length, void * private_data) { //struct vm_device * dev = (struct vm_device *)private_data; struct virtio_net_state * virtio = (struct virtio_net_state *)private_data; int port_idx = port % virtio->io_range_size; - PrintDebug("VIRTIO NIC %p Write for port %d (index=%d) len=%d, value=%x\n", private_data, port, port_idx, length, *(uint32_t *)src); - switch (port_idx) { case GUEST_FEATURES_PORT: if (length != 4) { PrintError("Illegal write length for guest features\n"); return -1; - } - + } virtio->virtio_cfg.guest_features = *(uint32_t *)src; PrintDebug("Setting Guest Features to %x\n", virtio->virtio_cfg.guest_features); - break; + case VRING_PG_NUM_PORT: - if (length != 4) { PrintError("Illegal write length for page frame number\n"); return -1; } - addr_t pfn = *(uint32_t *)src; addr_t page_addr = (pfn << VIRTIO_PAGE_SHIFT); uint16_t queue_idx = virtio->virtio_cfg.vring_queue_selector; - switch (queue_idx) { case 0: virtio_setup_queue(virtio, &virtio->rx_vq, pfn, page_addr); @@ -527,29 +545,24 @@ static int virtio_io_write(uint16_t port, void * src, uint_t length, void * priv break; case 2: virtio_setup_queue(virtio, &virtio->ctrl_vq, pfn, page_addr); - break; - + break; default: break; } - break; + case VRING_Q_SEL_PORT: virtio->virtio_cfg.vring_queue_selector = *(uint16_t *)src; - if (virtio->virtio_cfg.vring_queue_selector > 2) { PrintError("Virtio NIC device only uses 3 queue, selected %d\n", virtio->virtio_cfg.vring_queue_selector); return -1; } - break; + case VRING_Q_NOTIFY_PORT: { - uint16_t queue_idx = *(uint16_t *)src; - - //PrintDebug("Handling Kick\n"); - + uint16_t queue_idx = *(uint16_t *)src; if (queue_idx == 0){ PrintDebug("receive queue notification 0, packet get by Guest\n"); } else if (queue_idx == 1){ @@ -565,18 +578,16 @@ static int virtio_io_write(uint16_t port, void * src, uint_t length, void * priv } else { PrintError("Virtio NIC device only uses 3 queue, selected %d\n", queue_idx); - } - + } break; } + case VIRTIO_STATUS_PORT: virtio->virtio_cfg.status = *(uint8_t *)src; - if (virtio->virtio_cfg.status == 0) { PrintDebug("Resetting device\n"); - virtio_reset(virtio); + virtio_init_state(virtio); } - break; case VIRTIO_ISR_PORT: @@ -590,7 +601,6 @@ static int virtio_io_write(uint16_t port, void * src, uint_t length, void * priv return length; } - static int virtio_io_read(uint16_t port, void * dst, uint_t length, void * private_data) { //struct vm_device * dev = (struct vm_device *)private_data; @@ -607,19 +617,15 @@ static int virtio_io_read(uint16_t port, void * dst, uint_t length, void * priva PrintError("Illegal read length for host features\n"); return -1; } - *(uint32_t *)dst = virtio->virtio_cfg.host_features; - PrintDebug("value=0x%x\n", *(uint32_t *)dst); - break; - case VRING_PG_NUM_PORT: + case VRING_PG_NUM_PORT: if (length != 4) { PrintError("Illegal read length for page frame number\n"); return -1; } - switch (queue_idx) { case 0: *(uint32_t *)dst = virtio->rx_vq.pfn; @@ -633,17 +639,14 @@ static int virtio_io_read(uint16_t port, void * dst, uint_t length, void * priva default: break; } - PrintDebug(", value=0x%x\n", *(uint32_t *)dst); - break; - case VRING_SIZE_PORT: + case VRING_SIZE_PORT: if (length != 2) { PrintError("Illegal read length for vring size\n"); return -1; } - switch (queue_idx) { case 0: *(uint16_t *)dst = virtio->rx_vq.queue_size; @@ -657,29 +660,23 @@ static int virtio_io_read(uint16_t port, void * dst, uint_t length, void * priva default: break; } - PrintDebug("queue index: %d, value=0x%x\n", (int)queue_idx, *(uint16_t *)dst); - break; - case VIRTIO_STATUS_PORT: + case VIRTIO_STATUS_PORT: if (length != 1) { PrintError("Illegal read length for status\n"); return -1; } - *(uint8_t *)dst = virtio->virtio_cfg.status; - PrintDebug(", value=0x%x\n", *(uint8_t *)dst); break; - + case VIRTIO_ISR_PORT: *(uint8_t *)dst = virtio->virtio_cfg.pci_isr; virtio->virtio_cfg.pci_isr = 0; v3_pci_lower_irq(virtio->virtio_dev->pci_bus, 0, virtio->pci_dev); - - PrintDebug(", value=0x%x\n", *(uint8_t *)dst); - + PrintDebug(", value=0x%x\n", *(uint8_t *)dst); break; default: @@ -698,7 +695,6 @@ static struct v3_device_ops dev_ops = { .stop = NULL, }; - static int register_dev(struct virtio_dev_state * virtio, struct virtio_net_state * net_state) { struct pci_device * pci_dev = NULL; @@ -709,7 +705,6 @@ static int register_dev(struct virtio_dev_state * virtio, struct virtio_net_stat // This gets the number of ports, rounded up to a power of 2 net_state->io_range_size = 1; // must be a power of 2 - while (tmp_ports > 0) { tmp_ports >>= 1; net_state->io_range_size <<= 1; @@ -721,7 +716,6 @@ static int register_dev(struct virtio_dev_state * virtio, struct virtio_net_stat net_state->io_range_size >>= 1; } - for (i = 0; i < 6; i++) { bars[i].type = PCI_BAR_NONE; } @@ -731,7 +725,6 @@ static int register_dev(struct virtio_dev_state * virtio, struct virtio_net_stat bars[0].type = PCI_BAR_IO; bars[0].default_base_port = -1; bars[0].num_ports = net_state->io_range_size; - bars[0].io_read = virtio_io_read; bars[0].io_write = virtio_io_write; bars[0].private_data = net_state; @@ -754,34 +747,23 @@ static int register_dev(struct virtio_dev_state * virtio, struct virtio_net_stat pci_dev->config_header.device_id = VIRTIO_NET_DEV_ID; pci_dev->config_header.class = PCI_CLASS_NETWORK; - pci_dev->config_header.subclass = PCI_NET_SUBCLASS_OTHER; - - pci_dev->config_header.subsystem_id = VIRTIO_NET_SUBDEVICE_ID;; - + pci_dev->config_header.subclass = PCI_NET_SUBCLASS_OTHER; + pci_dev->config_header.subsystem_id = VIRTIO_NET_SUBDEVICE_ID; pci_dev->config_header.intr_pin = 1; - pci_dev->config_header.max_latency = 1; // ?? (qemu does it...) - - net_state->pci_dev = pci_dev; - //net_state->pci_bus = pci_bus; - + net_state->pci_dev = pci_dev; net_state->virtio_cfg.host_features = 0; //no features support now - net_state->rx_vq.queue_size = QUEUE_SIZE; net_state->tx_vq.queue_size = QUEUE_SIZE; net_state->ctrl_vq.queue_size = CTRL_QUEUE_SIZE; - net_state->virtio_dev = virtio; - net_state->pkt_sent = net_state->pkt_recv = net_state->pkt_drop = 0; - - virtio_reset(net_state); + + virtio_init_state(net_state); return 0; } - - static int connect_fn(struct guest_info * info, void * frontend_data, struct v3_dev_net_ops * ops, @@ -807,13 +789,8 @@ static int connect_fn(struct guest_info * info, //used when virtio_nic get a packet from guest and send it to the backend // send packet to all of the virtio nic devices other than the sender static int vnet_send(uint8_t * buf, uint32_t len, void * private_data, struct vm_device *dest_dev){ - PrintDebug("Virito NIC: In vnet_send: guest net state %p\n", private_data); -#ifdef CONFIG_DEBUG_VIRTIO_NET - print_packet(buf, len); -#endif - v3_vnet_send_rawpkt(buf, len, private_data); return 0; } @@ -827,10 +804,6 @@ static int vnet_receive(uint8_t * buf, uint32_t count, void * private_data, stru static int virtio_input(uchar_t * buf, uint_t len, void * private_data){ PrintDebug("Virito NIC: In virtio_input: guest net state %p\n", private_data); -#ifdef CONFIG_DEBUG_VIRTIO_NET - print_packet(buf, len); -#endif - return __virtio_dev_send(buf, len, private_data); } @@ -839,36 +812,35 @@ static int virtio_input(uchar_t * buf, uint_t len, void * private_data){ void register_virtio_to_vnet(struct vm_device *dev, char *dev_name, uchar_t mac[6]){ - struct virtio_net_state * net_state = (struct virtio_net_state *)V3_Malloc(sizeof(struct virtio_net_state)); - memset(net_state, 0, sizeof(struct virtio_net_state)); - + struct virtio_net_state * net_state; struct virtio_dev_state *virtio_state = (struct virtio_dev_state *)dev->private_data; + //uchar_t tapmac[6] = {0x00,0x02,0x55,0x67,0x42,0x39}; //for Intel-VT test HW + uchar_t tapmac[6] = {0x6e,0xa8,0x75,0xf4,0x82,0x95}; + uchar_t dstmac[6] = {0xff,0xff,0xff,0xff,0xff,0xff}; + uchar_t zeromac[6] = {0,0,0,0,0,0}; - net_state->net_ops = (struct v3_dev_net_ops *)V3_Malloc(sizeof(struct v3_dev_net_ops)); + net_state = (struct virtio_net_state *)V3_Malloc(sizeof(struct virtio_net_state)); + memset(net_state, 0, sizeof(struct virtio_net_state)); + net_state->net_ops = (struct v3_dev_net_ops *)V3_Malloc(sizeof(struct v3_dev_net_ops)); net_state->net_ops->send = &vnet_send; net_state->net_ops->receive = &vnet_receive; + net_state->dev = dev; register_dev(virtio_state, net_state); - - PrintDebug("Virtio NIC After register Device %s: queue size: %d, %d\n", dev->name, - net_state->rx_vq.queue_size, net_state->tx_vq.queue_size); - - PrintDebug("VNET: connect virtio nic state %p to vnet\n", net_state); - //add a device link to link table + PrintDebug("Virtio NIC register Device %s: queue size: %d, %d\n", dev->name, + net_state->rx_vq.queue_size, net_state->tx_vq.queue_size); + PrintDebug("Virtio NIC: connect virtio device %s, state %p, to vnet\n", dev->name, net_state); + int idx = vnet_register_device(dev, dev_name, mac, &virtio_input, net_state); - - uchar_t srcmac[6] = {0x00,0x02,0x55,0x67,0x42,0x39}; - uchar_t dstmac[6] = {0xff,0xff,0xff,0xff,0xff,0xff}; - uchar_t zeromac[6] = {0,0,0,0,0,0}; - vnet_add_route_entry(zeromac, dstmac, MAC_ANY, MAC_NONE, idx, LINK_INTERFACE, -1, LINK_INTERFACE); - if (idx == 0) - vnet_add_route_entry(zeromac, srcmac, MAC_ANY, MAC_NONE, idx, LINK_INTERFACE, -1, LINK_INTERFACE); - if (idx == 1) - vnet_add_route_entry(srcmac, zeromac, MAC_NONE, MAC_ANY, idx, LINK_INTERFACE, -1, LINK_INTERFACE); - + if (!strcmp(dev_name, "net_virtiodom0")) + vnet_add_route_entry(zeromac, tapmac, MAC_ANY, MAC_NONE, idx, LINK_INTERFACE, -1, LINK_INTERFACE); + if (!strcmp(dev_name, "net_virtio")) + vnet_add_route_entry(tapmac, zeromac, MAC_NONE, MAC_ANY, idx, LINK_INTERFACE, -1, LINK_INTERFACE); + + temp_net_states[net_idx ++] = net_state; } #endif @@ -903,8 +875,7 @@ static int virtio_init(struct guest_info * vm, v3_cfg_tree_t * cfg) { return -1; } - - //for temporary testing, add a backend + //for temporary testing, virtio bridge in Palacios #if 0 struct virtio_net_state * net_state = (struct virtio_net_state *)V3_Malloc(sizeof(struct virtio_net_state)); @@ -922,15 +893,21 @@ static int virtio_init(struct guest_info * vm, v3_cfg_tree_t * cfg) { temp_net_states[net_idx ++] = net_state; - PrintDebug("Net_states: 0: %p, 1: %p, 2: %p\n", temp_net_states[0], temp_net_states[1], temp_net_states[2]); + PrintDebug("Net_states: 0: %p, 1: %p, 2: %p\n", + temp_net_states[0], + temp_net_states[1], + temp_net_states[2]); #endif #if 1 //test interface between vnet & virtio-nic - uchar_t mac[6] = {0,0,0,0,0,0}; - register_virtio_to_vnet(dev, name,mac); + register_virtio_to_vnet(dev, name, mac); + PrintDebug("Net_states: 0: %p, 1: %p, 2: %p\n", + temp_net_states[0], + temp_net_states[1], + temp_net_states[2]); #endif return 0; diff --git a/palacios/src/palacios/svm_handler.c b/palacios/src/palacios/svm_handler.c index bd19242..75d2149 100644 --- a/palacios/src/palacios/svm_handler.c +++ b/palacios/src/palacios/svm_handler.c @@ -34,16 +34,28 @@ #include #include +#ifdef CONFIG_VNET +#include +#endif + +#ifdef CONFIG_LINUX_VIRTIO_NET + extern int v3_virtionic_pktprocess(struct guest_info * info); +#endif #ifdef CONFIG_TELEMETRY #include #endif -#ifdef CONFIG_VNET -#include +#define VNET_PROFILE +/* for vnet profiling*/ +#ifdef VNET_PROFILE +static uint64_t vmm_time = 0; +static uint64_t vnet_time = 0; +static uint64_t guest_time = 0; +static uint64_t last_exit_time = 0; +static uint64_t num_exit = 0; #endif - int v3_handle_svm_exit(struct guest_info * info, addr_t exit_code, addr_t exit_info1, addr_t exit_info2) { #ifdef CONFIG_TELEMETRY @@ -52,6 +64,14 @@ int v3_handle_svm_exit(struct guest_info * info, addr_t exit_code, addr_t exit_i } #endif +#ifdef VNET_PROFILE + uint64_t exit_start_time, vnet_start_time; + uint64_t exit_end_time, vnet_end_time; + rdtscll(exit_start_time); + num_exit ++; + if (last_exit_time > 0) + guest_time += exit_start_time - last_exit_time; +#endif //PrintDebug("SVM Returned: Exit Code: %x\n",exit_code); @@ -285,8 +305,34 @@ int v3_handle_svm_exit(struct guest_info * info, addr_t exit_code, addr_t exit_i } // END OF SWITCH (EXIT_CODE) + #ifdef CONFIG_VNET + +#ifdef VNET_PROFILE + rdtscll(vnet_start_time); +#endif v3_vnet_pkt_process(info); +#ifdef VNET_PROFILE + rdtscll(vnet_end_time); + vnet_time += vnet_end_time - vnet_start_time; +#endif +#endif + +#ifdef CONFIG_LINUX_VIRTIO_NET + v3_virtionic_pktprocess(info); +#endif + +#ifdef VNET_PROFILE + rdtscll(exit_end_time); + vmm_time += exit_end_time - exit_start_time; + last_exit_time = exit_end_time; + if ((num_exit % 100000) == 0) { + PrintError("exit: %ld, vmm_time: %ld, guest_time: %ld, vnet_time: %ld\n", (long)num_exit, (long)vmm_time, (long)guest_time, (long)vnet_time); + vmm_time = 0; + vnet_time = 0; + guest_time = 0; + last_exit_time = 0; + } #endif @@ -300,6 +346,7 @@ int v3_handle_svm_exit(struct guest_info * info, addr_t exit_code, addr_t exit_i if (exit_code == VMEXIT_INTR) { //PrintDebug("INTR ret IP = %x\n", guest_state->rip); } + return 0; } diff --git a/palacios/src/palacios/vmm.c b/palacios/src/palacios/vmm.c index c56356d..402cedf 100644 --- a/palacios/src/palacios/vmm.c +++ b/palacios/src/palacios/vmm.c @@ -25,6 +25,7 @@ #include #include #include +#include #ifdef CONFIG_SVM #include @@ -98,6 +99,10 @@ void Init_V3(struct v3_os_hooks * hooks, int num_cpus) { hooks->call_on_cpu(i, &init_cpu, (void *)(addr_t)i); } } + +#ifdef CONFIG_VNET + v3_init_vnet(); +#endif } diff --git a/palacios/src/palacios/vmm_config.c b/palacios/src/palacios/vmm_config.c index 002cf75..81a37d3 100644 --- a/palacios/src/palacios/vmm_config.c +++ b/palacios/src/palacios/vmm_config.c @@ -41,10 +41,6 @@ #endif -#ifdef CONFIG_VNET -#include -#endif - #include #include @@ -272,10 +268,6 @@ static int pre_config_guest(struct guest_info * info, struct v3_config * config_ v3_init_sym_swap(info); #endif -#ifdef CONFIG_VNET - v3_vnet_init(info); -#endif - if (schedule_hz_str) { sched_hz = atoi(schedule_hz_str); } diff --git a/palacios/src/palacios/vmm_queue.c b/palacios/src/palacios/vmm_queue.c index 2e1f72f..b06ff73 100644 --- a/palacios/src/palacios/vmm_queue.c +++ b/palacios/src/palacios/vmm_queue.c @@ -22,9 +22,9 @@ void v3_init_queue(struct gen_queue * queue) { queue->num_entries = 0; INIT_LIST_HEAD(&(queue->entries)); + v3_lock_init(&queue->lock); } - struct gen_queue * v3_create_queue() { struct gen_queue * tmp_queue = V3_Malloc(sizeof(struct gen_queue)); v3_init_queue(tmp_queue); @@ -34,15 +34,18 @@ struct gen_queue * v3_create_queue() { void v3_enqueue(struct gen_queue * queue, addr_t entry) { struct queue_entry * q_entry = V3_Malloc(sizeof(struct queue_entry)); + v3_lock(queue->lock); q_entry->entry = entry; list_add_tail(&(q_entry->entry_list), &(queue->entries)); queue->num_entries++; + v3_unlock(queue->lock); } addr_t v3_dequeue(struct gen_queue * queue) { addr_t entry_val = 0; + v3_lock(queue->lock); if (!list_empty(&(queue->entries))) { struct list_head * q_entry = queue->entries.next; struct queue_entry * tmp_entry = list_entry(q_entry, struct queue_entry, entry_list); @@ -51,6 +54,7 @@ addr_t v3_dequeue(struct gen_queue * queue) { list_del(q_entry); V3_Free(tmp_entry); } + v3_unlock(queue->lock); return entry_val; } diff --git a/palacios/src/palacios/vmm_vnet.c b/palacios/src/palacios/vmm_vnet.c index 188f730..dffb6d4 100644 --- a/palacios/src/palacios/vmm_vnet.c +++ b/palacios/src/palacios/vmm_vnet.c @@ -9,13 +9,11 @@ * * Copyright (c) 2009, Lei Xia * Copyright (c) 2009, Yuan Tang - * Copyright (c) 2009, Zheng Cui * Copyright (c) 2009, The V3VEE Project * All rights reserved. * * Author: Lei Xia * Yuan Tang - * Zheng Cui * * This is free software. You are permitted to use, * redistribute, and modify it as specified in the file "V3VEE_LICENSE". @@ -24,21 +22,13 @@ #include #include #include +#include #ifndef CONFIG_DEBUG_VNET #undef PrintDebug #define PrintDebug(fmt, args...) #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]; -}__attribute__((packed)); - - // 14 (ethernet frame) + 20 bytes struct in_pkt_header { uchar_t ethernetdest[6]; @@ -47,44 +37,40 @@ struct in_pkt_header { char ip[20]; }__attribute__((packed)); -#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; +#define HASH_KEY_SIZE 16 struct link_table { struct link_entry * links[MAX_LINKS]; uint16_t size; + v3_lock_t lock; }__attribute__((packed)); struct routing_table { struct routing_entry * routes[MAX_ROUTES]; uint16_t size; + v3_lock_t lock; }__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; -// This is the hash value, Format: 0: num_matched_routes, 1...n: matches[] -- TY struct route_cache_entry { int num_matched_routes; int * matches; }; -// the route cache -static struct hashtable * g_route_cache; +struct vnet_state_t { + struct link_table g_links; + struct routing_table g_routes; + struct gen_queue * g_inpkt_q; + struct hashtable * g_route_cache; + v3_lock_t cache_lock; +}; + +static struct vnet_state_t g_vnet_state;//global state for vnet static uint16_t ip_xsum(struct ip_header *ip_hdr, int hdr_len){ long sum = 0; @@ -106,27 +92,30 @@ static uint16_t ip_xsum(struct ip_header *ip_hdr, int hdr_len){ return (uint16_t)~sum; } -static inline void ethernet_packet_init(struct ethernet_pkt *pt, uchar_t *data, const size_t size) { +static void ethernet_packet_init(struct ethernet_pkt *pt, + uchar_t *data, + const size_t size) { pt->size = size; + pt->use_header = 0; + memset(&pt->ext_hdr, 0, sizeof(struct udp_link_header)); memcpy(pt->data, data, size); } -static uint_t hash_from_key_fn(addr_t hashkey) { - uint8_t * key = (uint8_t *)hashkey; - return v3_hash_buffer(key, HASH_KEY_LEN); +static inline uint_t hash_from_key_fn(addr_t hashkey) { + return v3_hash_buffer((uint8_t *)hashkey, HASH_KEY_LEN); } -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 inline int hash_key_equal(addr_t key1, addr_t key2) { + return (memcmp((uint8_t *)key1, (uint8_t *)key2, 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); +static int init_route_cache(struct vnet_state_t *vnet_state) { + vnet_state->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"); + if (vnet_state->g_route_cache == NULL) { + PrintError("Vnet: Route Cache Init Fails\n"); return -1; } @@ -134,10 +123,10 @@ static int init_route_cache() { } static void make_hash_key(route_hashkey_t hashkey, - char src_addr[6], - char dest_addr[6], - char src_type, - int src_index) { + char src_addr[6], + char dest_addr[6], + char src_type, + int src_index) { int j; for (j = 0; j < 6; j++) { @@ -150,8 +139,11 @@ static void make_hash_key(route_hashkey_t hashkey, *(int *)(hashkey + 12) = src_index; } -static int add_route_to_cache(route_hashkey_t hashkey, int num_matched_r, int * matches) { +static int add_route_to_cache(route_hashkey_t hashkey, + int num_matched_r, + int * matches) { struct route_cache_entry * new_entry = NULL; + struct vnet_state_t *vnet_state = &g_vnet_state; int i; new_entry = (struct route_cache_entry *)V3_Malloc(sizeof(struct route_cache_entry)); @@ -161,7 +153,6 @@ static int add_route_to_cache(route_hashkey_t hashkey, int num_matched_r, int * } new_entry->num_matched_routes = num_matched_r; - new_entry->matches = (int *)V3_Malloc(num_matched_r * sizeof(int)); if (new_entry->matches == NULL) { @@ -173,9 +164,8 @@ static int add_route_to_cache(route_hashkey_t hashkey, int num_matched_r, int * 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"); + if (v3_htable_insert(vnet_state->g_route_cache, (addr_t)hashkey, (addr_t)new_entry) == 0) { + PrintError("Vnet: Failed to insert new route entry to the cache\n"); V3_Free(new_entry->matches); V3_Free(new_entry); } @@ -184,11 +174,14 @@ static int add_route_to_cache(route_hashkey_t hashkey, int num_matched_r, int * } 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); + struct vnet_state_t *vnet_state = &g_vnet_state; + + v3_free_htable(vnet_state->g_route_cache, 1, 1); + vnet_state->g_route_cache = v3_create_htable(MIN_CACHE_SIZE, + hash_from_key_fn, + hash_key_equal); - if (g_route_cache == NULL) { + if (vnet_state->g_route_cache == NULL) { PrintError("Vnet: Route Cache Create Failurely\n"); return -1; } @@ -200,8 +193,10 @@ static int look_into_cache(route_hashkey_t hashkey, int * matches) { struct route_cache_entry * found = NULL; int n_matches = -1; int i = 0; + struct vnet_state_t *vnet_state = &g_vnet_state; - found = (struct route_cache_entry *)v3_htable_search(g_route_cache, (addr_t)hashkey); + found = (struct route_cache_entry *)v3_htable_search(vnet_state->g_route_cache, + (addr_t)hashkey); if (found != NULL) { n_matches = found->num_matched_routes; @@ -218,7 +213,7 @@ static int look_into_cache(route_hashkey_t hashkey, int * matches) { #ifdef CONFIG_DEBUG_VNET static void print_packet(uchar_t *pkt, int size) { - PrintDebug("Vnet: print_data_packet: size: %d\n", size); + PrintDebug("Vnet: data_packet: size: %d\n", size); v3_hexdump(pkt, size, NULL, 0); } @@ -251,24 +246,35 @@ static inline void mac_to_string(char mac[6], char * buf) { mac[3], mac[4], mac[5]); } -static void dump_routes(struct routing_entry **route_tables) { + +static void print_route(struct routing_entry *route){ char dest_str[18]; char src_str[18]; - struct routing_entry *route = NULL; + + mac_to_string(route->src_mac, src_str); + mac_to_string(route->dest_mac, dest_str); + + PrintDebug("SRC(%s), DEST(%s), src_mac_qual(%d), dst_mac_qual(%d)\n", + src_str, + dest_str, + route->src_mac_qual, + route->dest_mac_qual); + PrintDebug("Src_Link(%d), src_type(%d), dst_link(%d), dst_type(%d)\n\n", + route->src_link_idx, + route->src_type, + route->link_idx, + route->link_type); +} + + +static void dump_routes(struct routing_entry **route_tables) { int i; PrintDebug("\nVnet: route table dump start =====\n"); for(i = 0; i < MAX_ROUTES; i++) { if (route_tables[i] != NULL){ - route = route_tables[i]; - - mac_to_string(route->src_mac, src_str); - mac_to_string(route->dest_mac, dest_str); - - PrintDebug("route: %d\n", i); - PrintDebug("SRC(%s), DEST(%s), src_mac_qual(%d), dst_mac_qual(%d)\n", src_str, dest_str, route->src_mac_qual, route->dest_mac_qual); - PrintDebug("Src_Link(%d), src_type(%d), dst_link(%d), dst_type(%d)\n\n", route->src_link_idx, route->src_type, route->link_idx, route->link_type); + print_route(route_tables[i]); } } @@ -276,20 +282,12 @@ static void dump_routes(struct routing_entry **route_tables) { } static void dump_dom0_routes(struct routing_entry routes[], int size) { - char dest_str[18]; - char src_str[18]; - struct routing_entry *route = NULL; int i; PrintDebug("\nVnet: route table from dom0 guest =====\n"); for(i = 0; i src_mac, src_str); - mac_to_string(route->dest_mac, dest_str); - PrintDebug("route: %d\n", i); - PrintDebug("SRC(%s), DEST(%s), src_mac_qual(%d), dst_mac_qual(%d)\n", src_str, dest_str, route->src_mac_qual, route->dest_mac_qual); - PrintDebug("Src_Link(%d), src_type(%d), dst_link(%d), dst_type(%d)\n\n", route->src_link_idx, route->src_type, route->link_idx, route->link_type); + print_route(&routes[i]); } PrintDebug("\nVnet: route table dom0 guest end =====\n"); @@ -304,9 +302,12 @@ static void dump_dom0_links(struct vnet_if_link links[], int size) { for(i = 0; i dest_ip, link->dest_port, link->pro_type); + PrintDebug("dest_ip(%ld), dst_port(%d), prot_type(%d)\n", + link->dest_ip, + link->dest_port, + link->pro_type); PrintDebug("vnet_header:\n"); - v3_hexdump(&link->vnet_header, sizeof(struct udp_link_header), NULL, 0); + v3_hexdump(&link->vnet_header, sizeof(struct udp_link_header), NULL, 0); } PrintDebug("\nVnet: link table dom0 guest end =====\n"); @@ -316,36 +317,51 @@ static void dump_dom0_links(struct vnet_if_link links[], int size) { static int __add_link_entry(struct link_entry * link) { int idx; - + struct vnet_state_t *vnet_state = &g_vnet_state; + + v3_lock(vnet_state->g_links.lock); for (idx = 0; idx < MAX_LINKS; idx++) { - if (g_links.links[idx] == NULL) { - g_links.links[idx] = link; - g_links.size++; - - return idx; + if (vnet_state->g_links.links[idx] == NULL) { + vnet_state->g_links.links[idx] = link; + vnet_state->g_links.size++; + break; } } + v3_unlock(vnet_state->g_links.lock); - PrintError("No available Link entry\n"); - return -1; + if (idx == MAX_LINKS) { + PrintDebug("VNET: No available Link entry for new link\n"); + return -1; + } + + return idx; } static int __add_route_entry(struct routing_entry * route) { int idx; - + struct vnet_state_t *vnet_state = &g_vnet_state; + + v3_lock(vnet_state->g_routes.lock); for (idx = 0; idx < MAX_ROUTES; idx++) { - if (g_routes.routes[idx] == NULL) { - g_routes.routes[idx] = route; - g_routes.size++; - - return idx; + if (vnet_state->g_routes.routes[idx] == NULL) { + vnet_state->g_routes.routes[idx] = route; + vnet_state->g_routes.size++; + break; } } + v3_unlock(vnet_state->g_routes.lock); - PrintError("No available route entry\n"); - return -1; -} + if(idx == MAX_LINKS){ + PrintDebug("VNET: No available route entry for new route\n"); + return -1; + } +#ifdef CONFIG_DEBUG_VNET + dump_routes(vnet_state->g_routes.routes); +#endif + + return idx; +} int vnet_add_route_entry(char src_mac[6], char dest_mac[6], @@ -358,8 +374,9 @@ int vnet_add_route_entry(char src_mac[6], struct routing_entry * new_route = (struct routing_entry *)V3_Malloc(sizeof(struct routing_entry)); int idx = -1; + PrintDebug("Vnet: vnet_add_route_entry\n"); + memset(new_route, 0, sizeof(struct routing_entry)); - if ((src_mac_qual != MAC_ANY)) { memcpy(new_route->src_mac, src_mac, 6); } @@ -381,143 +398,33 @@ int vnet_add_route_entry(char src_mac[6], } clear_hash_cache(); - - return idx; -} -#if 0 -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; - } - - g_links.links[index] = NULL; - g_links.size--; - - V3_Free(link); - - return ret; -} - -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; -} - -static int __delete_route_entry(int index) { - struct routing_entry * route; - - if ((index >= MAX_ROUTES) || (g_routes.routes[index] == NULL)) { - PrintDebug("VNET: wrong index in delete route entry %d\n", index); - return -1; - } - - route = g_routes.routes[index]; - g_routes.routes[index] = NULL; - g_routes.size--; - - V3_Free(route); - - clear_hash_cache(); - - return 0; -} - -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; - } - - return __delete_route_entry(index); + return idx; } -#endif static int match_route(uint8_t * src_mac, uint8_t * dst_mac, link_type_t src_type, int src_index, int * matches) { + struct routing_entry *route = NULL; + struct vnet_state_t *vnet_state = &g_vnet_state; int matched_routes[MAX_ROUTES]; int num_matches = 0; int i; - struct routing_entry *route = NULL; #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: match_route. pkt: SRC(%s), DEST(%s)\n", src_str, dest_str); - - dump_routes(g_routes.routes); #endif for(i = 0; i < MAX_ROUTES; i++) { - if (g_routes.routes[i] != NULL){ - route = g_routes.routes[i]; + if (vnet_state->g_routes.routes[i] != NULL){ + route = vnet_state->g_routes.routes[i]; if(src_type == LINK_ANY && src_index == -1) { if ((route->dest_mac_qual == MAC_ANY) && @@ -590,17 +497,15 @@ static int match_route(uint8_t * src_mac, 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]; - int matches[MAX_ROUTES]; int num_matched_routes = 0; - struct in_pkt_header header; + char hash_key[HASH_KEY_SIZE]; + struct vnet_state_t *vnet_state = &g_vnet_state; + int i; - 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); @@ -609,22 +514,17 @@ static int handle_one_pkt(struct ethernet_pkt *pkt) { #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 - // 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); - + 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); } @@ -643,14 +543,14 @@ static int handle_one_pkt(struct ethernet_pkt *pkt) { struct link_entry * link = NULL; route_index = matches[i]; - link_index = g_routes.routes[route_index]->link_idx; + link_index = vnet_state->g_routes.routes[route_index]->link_idx; if ((link_index < 0) || (link_index > MAX_LINKS) || - (g_links.links[link_index] == NULL)) { + (vnet_state->g_links.links[link_index] == NULL)) { continue; } - link = g_links.links[link_index]; + link = vnet_state->g_links.links[link_index]; pkt_len = pkt->size; if (link->type == LINK_EDGE) { @@ -659,9 +559,9 @@ static int handle_one_pkt(struct ethernet_pkt *pkt) { 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)); + 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); @@ -691,52 +591,54 @@ static int handle_one_pkt(struct ethernet_pkt *pkt) { return 0; } -static int send_ethernet_pkt(uchar_t *data, int len) { +static int send_ethernet_pkt(uchar_t *data, int len, void *private_data) { struct ethernet_pkt *pkt; + struct vnet_state_t *vnet_state = &g_vnet_state; pkt = (struct ethernet_pkt *)V3_Malloc(sizeof(struct ethernet_pkt)); - if(pkt == NULL){ PrintError("VNET: Memory allocate fails\n"); return -1; } - memset(pkt, 0, sizeof(struct ethernet_pkt)); - - 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); + memset(pkt, 0, sizeof(struct ethernet_pkt)); + ethernet_packet_init(pkt, data, len); + v3_enqueue(vnet_state->g_inpkt_q, (addr_t)pkt); + #ifdef CONFIG_DEBUG_VNET + PrintDebug("VNET: send_pkt: transmitting packet: (size:%d)\n", (int)pkt->size); print_packet((char *)data, len); #endif - - 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); +int v3_vnet_send_rawpkt(uchar_t * buf, + int len, + void *private_data) { + PrintDebug("VNET: In v3_vnet_send_rawpkt: pkt length %d\n", len); - return send_ethernet_pkt(buf, len); + return send_ethernet_pkt(buf, len, private_data); } //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); - +int v3_vnet_send_udppkt(uchar_t * buf, + int len, + void *private_data) { uint_t hdr_len = sizeof(struct udp_link_header); + + PrintDebug("VNET: In v3_vnet_send_udppkt: pkt length %d\n", len); - return send_ethernet_pkt((uchar_t *)(buf+hdr_len), len - hdr_len); + return send_ethernet_pkt((uchar_t *)(buf+hdr_len), len - hdr_len, private_data); } static int search_device(char * device_name) { + struct vnet_state_t *vnet_state = &g_vnet_state; 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) { + if ((vnet_state->g_links.links[i] != NULL) && (vnet_state->g_links.links[i]->type == LINK_INTERFACE)) { + if (strcmp(device_name, vnet_state->g_links.links[i]->dst_dev->name) == 0) { return i; } } @@ -779,38 +681,13 @@ int vnet_register_device(struct vm_device * vdev, return idx; } -#if 0 -static int vnet_unregister_device(char * dev_name) { - int idx; - - idx = search_device(dev_name); - - if (idx == -1) { - PrintDebug("VNET: No device with name %s found\n", dev_name); - return -1; - } - - 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); - return -1; - } - - V3_Free(device); - - return idx; -} - -#endif - int v3_vnet_pkt_process() { struct ethernet_pkt * pkt; + struct vnet_state_t *vnet_state = &g_vnet_state; - 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); - + while ((pkt = (struct ethernet_pkt *)v3_dequeue(vnet_state->g_inpkt_q))!= NULL) { if (handle_one_pkt(pkt) != -1) { - PrintDebug("VNET: vnet_check: handle one packet!\n"); + PrintDebug("VNET: vnet_check: handle one packet! pt length %d, pt type %d\n", (int)pkt->size, (int)pkt->type); } else { PrintDebug("VNET: vnet_check: Fail to forward one packet, discard it!\n"); } @@ -821,37 +698,36 @@ int v3_vnet_pkt_process() { return 0; } -static void init_empty_link_table() { +static void vnet_state_init(struct vnet_state_t *vnet_state) { int i; + /*initial links table */ for (i = 0; i < MAX_LINKS; i++) { - g_links.links[i] = NULL; + vnet_state->g_links.links[i] = NULL; } + vnet_state->g_links.size = 0; + if(v3_lock_init(&(vnet_state->g_links.lock)) == -1){ + PrintError("VNET: Failure to init lock for links table\n"); + } + PrintDebug("VNET: Links table initiated\n"); - g_links.size = 0; -} - -static void init_empty_route_table() { - int i; - + /*initial routes table */ for (i = 0; i < MAX_ROUTES; i++) { - g_routes.routes[i] = NULL; + vnet_state->g_routes.routes[i] = NULL; } + vnet_state->g_routes.size = 0; + if(v3_lock_init(&(vnet_state->g_routes.lock)) == -1){ + PrintError("VNET: Failure to init lock for routes table\n"); + } + PrintDebug("VNET: Routes table initiated\n"); - g_links.size = 0; -} - -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"); - - g_inpkt_q = v3_create_queue(); - v3_init_queue(g_inpkt_q); + /*initial pkt receiving queue */ + vnet_state->g_inpkt_q = v3_create_queue(); + v3_init_queue(vnet_state->g_inpkt_q); + PrintDebug("VNET: Receiving queue initiated\n"); + + /*initial routing cache */ + init_route_cache(vnet_state); } static void free_link_mem(struct link_entry *link){ @@ -859,7 +735,8 @@ static void free_link_mem(struct link_entry *link){ V3_Free(link); } -// TODO: + +// TODO: static int addto_routing_link_tables(struct routing_entry *route_tab, uint16_t num_routes, struct link_entry *link_tab, @@ -943,15 +820,13 @@ static int addto_routing_link_tables(struct routing_entry *route_tab, } new_route = (struct routing_entry *)V3_Malloc(sizeof(struct routing_entry)); - 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]; - + 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]; @@ -1009,13 +884,18 @@ static int handle_init_tables_hcall(struct guest_info * info, uint_t hcall_id, v return 0; } -void v3_vnet_init(struct guest_info * vm) { - init_tables(); - init_pkt_queue(); +void v3_init_vnet() { + vnet_state_init(&g_vnet_state); - v3_register_hypercall(vm, VNET_INITAB_HCALL, handle_init_tables_hcall, NULL); - PrintDebug("VNET Initialized\n"); } +//only need to called in config dom0 guest +int v3_register_guest_vnet(struct guest_info *vm){ + return v3_register_hypercall(vm, + VNET_INITAB_HCALL, + handle_init_tables_hcall, + NULL); +} + diff --git a/palacios/src/palacios/vmx_handler.c b/palacios/src/palacios/vmx_handler.c index 67f03a4..c758abf 100644 --- a/palacios/src/palacios/vmx_handler.c +++ b/palacios/src/palacios/vmx_handler.c @@ -32,15 +32,27 @@ #include #include -#ifdef CONFIG_TELEMETRY -#include +#ifdef CONFIG_VNET +#include #endif +#ifdef CONFIG_LINUX_VIRTIO_NET + extern int v3_virtionic_pktprocess(struct guest_info * info); +#endif -#ifdef CONFIG_VNET -#include +#ifdef CONFIG_TELEMETRY +#include #endif +//#define VNET_PROFILE +/* for vnet profiling*/ +#ifdef VNET_PROFILE +static uint64_t vmm_time = 0; +static uint64_t vnet_time = 0; +static uint64_t guest_time = 0; +static uint64_t last_exit_time = 0; +static uint64_t num_exit = 0; +#endif /* At this point the GPRs are already copied into the guest_info state */ @@ -61,6 +73,15 @@ int v3_handle_vmx_exit(struct guest_info * info, struct vmx_exit_info * exit_inf } #endif +#ifdef VNET_PROFILE + uint64_t exit_start_time, vnet_start_time; + uint64_t exit_end_time, vnet_end_time; + rdtscll(exit_start_time); + num_exit ++; + if (last_exit_time > 0) + guest_time += exit_start_time - last_exit_time; +#endif + switch (exit_info->exit_reason) { case VMEXIT_INFO_EXCEPTION_OR_NMI: { pf_error_t error_code = *(pf_error_t *)&(exit_info->int_err); @@ -226,16 +247,44 @@ int v3_handle_vmx_exit(struct guest_info * info, struct vmx_exit_info * exit_inf return -1; } + #ifdef CONFIG_VNET +#ifdef VNET_PROFILE + rdtscll(vnet_start_time); +#endif v3_vnet_pkt_process(info); +#ifdef VNET_PROFILE + rdtscll(vnet_end_time); + vnet_time += vnet_end_time - vnet_start_time; +#endif #endif +#ifdef CONFIG_LINUX_VIRTIO_NET + v3_virtionic_pktprocess(info); +#endif + +#ifdef VNET_PROFILE + rdtscll(exit_end_time); + vmm_time += exit_end_time - exit_start_time; + last_exit_time = exit_end_time; + if ((num_exit % 100000) == 0) { + PrintError("exit: %ld, vmm_time: %ld, guest_time: %ld, vnet_time: %ld\n", (long)num_exit, (long)vmm_time, (long)guest_time, (long)vnet_time); + vmm_time = 0; + vnet_time = 0; + guest_time = 0; + last_exit_time = 0; + } +#endif + + + #ifdef CONFIG_TELEMETRY if (info->enable_telemetry) { v3_telemetry_end_exit(info, exit_info->exit_reason); } #endif + return 0; } diff --git a/utils/guest_creator/default.xml b/utils/guest_creator/default.xml index d8576e7..ee1b098 100644 --- a/utils/guest_creator/default.xml +++ b/utils/guest_creator/default.xml @@ -27,7 +27,7 @@ - + @@ -108,12 +108,12 @@ 0 - +