X-Git-Url: http://v3vee.org/palacios/gitweb/gitweb.cgi?p=palacios.git;a=blobdiff_plain;f=palacios%2Fsrc%2Fdevices%2Flnx_virtio_nic.c;h=e83a1992505885a4bf54a6fb5033eac8565c0a9b;hp=317837dd9ef37614d4769db3ee13880bce430875;hb=68521eca0f94f8c22aa60e762c83e24beec7233c;hpb=9f239e337227351e9440d9d3b174bb37942ceebe diff --git a/palacios/src/devices/lnx_virtio_nic.c b/palacios/src/devices/lnx_virtio_nic.c index 317837d..e83a199 100644 --- a/palacios/src/devices/lnx_virtio_nic.c +++ b/palacios/src/devices/lnx_virtio_nic.c @@ -7,13 +7,13 @@ * and the University of New Mexico. You can find out more at * http://www.v3vee.org * - * Copyright (c) 2008, Jack Lange * Copyright (c) 2008, Lei Xia + * Copyright (c) 2008, Cui Zheng * Copyright (c) 2008, The V3VEE Project * All rights reserved. * - * Author: Jack Lange - * Lei Xia + * Author: Lei Xia + * Cui Zheng * * * This is free software. You are permitted to use, @@ -23,58 +23,29 @@ #include #include #include -#include #include +#include +#include +#include #include -#ifndef CONFIG_DEBUG_VIRTIO_BLK +#ifndef CONFIG_DEBUG_VIRTIO_NET #undef PrintDebug #define PrintDebug(fmt, args...) #endif -#define NIC_STATUS_OK 0 -#define NIC_STATUS_ERR 1 -#define NIC_STATUS_NOT_SUPPORTED 2 - - -/* 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 */ +//#define VIRTIO_NIC_PROFILE -/* Maximum packet size we can receive from tap device: header + 64k */ +#define VIRTIO_NET_S_LINK_UP 1 /* Link is up */ #define VIRTIO_NET_MAX_BUFSIZE (sizeof(struct virtio_net_hdr) + (64 << 10)) - struct virtio_net_hdr { -#define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 /* Use csum_start, csum_offset */ uint8_t flags; #define VIRTIO_NET_HDR_GSO_NONE 0 /* Not a GSO frame */ -#define VIRTIO_NET_HDR_GSO_TCPV4 1 /* GSO frame, IPv4 TCP (TSO) */ -#define VIRTIO_NET_HDR_GSO_UDP 3 /* GSO frame, IPv4 UDP (UFO) */ -#define VIRTIO_NET_HDR_GSO_TCPV6 4 /* GSO frame, IPv6 TCP */ -#define VIRTIO_NET_HDR_GSO_ECN 0x80 /* TCP has ECN set */ 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 */ @@ -82,45 +53,89 @@ struct virtio_net_hdr { }__attribute__((packed)); - - -#define QUEUE_SIZE 256 +#define QUEUE_SIZE 4096 #define CTRL_QUEUE_SIZE 64 - - -struct v3_net_ops { - int (*send)(uint8_t * buf, uint32_t count, void * private_data); - int (*receive)(uint8_t * buf, uint32_t count, void * private_data); -}; - - #define ETH_ALEN 6 struct virtio_net_config { - uint8_t mac[ETH_ALEN]; - // See VIRTIO_NET_F_STATUS and VIRTIO_NET_S_* above + uint8_t mac[ETH_ALEN]; //VIRTIO_NET_F_MAC uint16_t status; } __attribute__((packed)); +struct virtio_dev_state { + struct vm_device * pci_bus; + struct list_head dev_list; + struct guest_info * vm; +}; + struct virtio_net_state { struct virtio_net_config net_cfg; struct virtio_config virtio_cfg; - struct vm_device * pci_bus; - struct pci_device * pci_dev; + struct vm_device * dev; + struct pci_device * pci_dev; + int io_range_size; struct virtio_queue rx_vq; //index 0, rvq in Linux virtio driver, handle packet to guest 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 - struct v3_net_ops * net_ops; + ulong_t pkt_sent, pkt_recv, pkt_drop; + struct gen_queue * inpkt_q; - int io_range_size; + struct v3_dev_net_ops * net_ops; - void *private_data; + void * backend_data; + struct virtio_dev_state * virtio_dev; + struct list_head dev_link; }; +struct eth_pkt { + uint32_t size; //size of data + char data[ETHERNET_PACKET_LEN]; +}__attribute__((packed)); + +//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 + +static struct virtio_net_state *temp_net_states[3]; +static int net_idx = 0; + +static int __virtio_dev_send(uchar_t * buf, uint32_t size, void *private_data); + +static int send(uint8_t * buf, uint32_t count, void * private_data, struct vm_device *dest_dev) +{ + PrintDebug("Virito NIC: In sending stub, guest %p, count %d\n", private_data, count); + +#ifdef CONFIG_DEBUG_VIRTIO_NET + print_packet(buf, 20); +#endif + + struct virtio_net_state *virtio_state = (struct virtio_net_state *)private_data; + + if (virtio_state == temp_net_states[0]) + __virtio_dev_send(buf, count, temp_net_states[1]); + + if (virtio_state == temp_net_states[1]){ //return a RARP packet + __virtio_dev_send(buf, count, temp_net_states[0]); + } + + return count; +} + +static int receive(uint8_t * buf, uint32_t count, void * private_data, struct vm_device *src_dev) +{ + PrintDebug("Virito NIC: In receiveing stub\n"); + + return 0; +} + +#endif static int virtio_free(struct vm_device * dev) { @@ -128,10 +143,8 @@ static int virtio_free(struct vm_device * dev) return -1; } -static int virtio_reset(struct vm_device * dev) +static int virtio_init_state(struct virtio_net_state * virtio) { - struct virtio_net_state * virtio = (struct virtio_net_state *)dev->private_data; - virtio->rx_vq.ring_desc_addr = 0; virtio->rx_vq.ring_avail_addr = 0; virtio->rx_vq.ring_used_addr = 0; @@ -150,98 +163,236 @@ static int virtio_reset(struct vm_device * dev) virtio->ctrl_vq.pfn = 0; virtio->ctrl_vq.cur_avail_idx = 0; - virtio->virtio_cfg.status = VIRTIO_NET_S_LINK_UP; + virtio->virtio_cfg.host_features = 0; + //virtio->virtio_cfg.status = VIRTIO_NET_S_LINK_UP; virtio->virtio_cfg.pci_isr = 0; - virtio->private_data = NULL; + + virtio->pkt_sent = virtio->pkt_recv = virtio->pkt_drop = 0; + + virtio->inpkt_q = v3_create_queue(); + v3_init_queue(virtio->inpkt_q); return 0; } -static int read_op(struct vm_device * dev, uint8_t * buf, uint32_t len) + +//sending guest's packet to network sink +static int pkt_write(struct virtio_net_state * virtio, struct vring_desc * buf_desc) { - struct virtio_net_state * virtio = (struct virtio_net_state *)dev->private_data; - int ret = -1; + //struct virtio_net_state * virtio = (struct virtio_net_state *)dev->private_data; + uint8_t * buf = NULL; + uint32_t len = buf_desc->length; - PrintDebug("Receving pkt from guest\n"); + PrintDebug("Virtio NIC: Handling Virtio Write, net_state: %p\n", virtio); - ret = virtio->net_ops->receive(buf, len, virtio->private_data); - return ret; -} + if (guest_pa_to_host_va(virtio->virtio_dev->vm, buf_desc->addr_gpa, (addr_t *)&(buf)) == -1) { + PrintError("Could not translate buffer address\n"); + return -1; + } + if (virtio->net_ops->send(buf, len, (void *)virtio, NULL) == -1) { + return -1; + } -static int write_op(struct vm_device *dev, uint8_t *buf, uint32_t len) -{ - struct virtio_net_state * virtio = (struct virtio_net_state *)dev->private_data; - int ret = -1; + return 0; +} - PrintDebug("Receving pkt from guest\n"); - ret = virtio->net_ops->send(buf, len, virtio->private_data); +static int build_receive_header(struct virtio_net_hdr * hdr, const void * buf, int raw) { + hdr->flags = 0; + + if (!raw) { + memcpy(hdr, buf, sizeof(struct virtio_net_hdr)); + } else { + memset(hdr, 0, sizeof(struct virtio_net_hdr)); + } - return ret; + return 0; } //sending guest's packet to network sink -static int handle_pkt_write(struct vm_device *dev, struct virtio_net_hdr *hdr, - struct vring_desc *buf_desc, uint8_t *status) +static int copy_data_to_desc(struct virtio_net_state * virtio_state, struct vring_desc * desc, uchar_t * buf, uint_t buf_len) { - //struct virtio_net_state * virtio = (struct virtio_net_state *)dev->private_data; - uint8_t * buf = NULL; - - PrintDebug("Handling Virtio Net write\n"); + uint32_t len; + uint8_t * desc_buf = NULL; - if (guest_pa_to_host_va(dev->vm, buf_desc->addr_gpa, (addr_t *)&(buf)) == -1) { + if (guest_pa_to_host_va(virtio_state->virtio_dev->vm, desc->addr_gpa, (addr_t *)&(desc_buf)) == -1) { PrintError("Could not translate buffer address\n"); return -1; } + len = (desc->length < buf_len)?desc->length:buf_len; + memcpy(desc_buf, buf, len); + + return len; +} + + +//send data to guest +static int send_pkt_to_guest(struct virtio_net_state * virtio, uchar_t * buf, uint_t size, int raw, void * private_data) +{ + struct virtio_queue * q = &(virtio->rx_vq); + struct virtio_net_hdr hdr; + uint32_t hdr_len = sizeof(struct virtio_net_hdr); + uint32_t data_len = size; + uint32_t offset = 0; + + PrintDebug("VIRTIO NIC: sending packet to virtio nic %p, size:%d", virtio, size); + + virtio->pkt_recv ++; + if (!raw) { + data_len -= hdr_len; + } - PrintDebug("Length=%d\n", buf_desc->length); + build_receive_header(&hdr, buf, 1); - if (write_op(dev, buf, buf_desc->length) == -1) { - *status = NIC_STATUS_ERR; + if (q->ring_avail_addr == 0) { + PrintError("Queue is not set\n"); return -1; + } + + if (q->last_avail_idx > q->avail->index) + q->idx_overflow = true; + q->last_avail_idx = q->avail->index; + + if (q->cur_avail_idx < q->avail->index || (q->idx_overflow && q->cur_avail_idx < q->avail->index+65536)){ + addr_t hdr_addr = 0; + uint16_t hdr_idx = q->avail->ring[q->cur_avail_idx % q->queue_size]; + uint16_t buf_idx = 0; + struct vring_desc * hdr_desc = NULL; + + hdr_desc = &(q->desc[hdr_idx]); + if (guest_pa_to_host_va(virtio->virtio_dev->vm, hdr_desc->addr_gpa, &(hdr_addr)) == -1) { + PrintError("Could not translate receive buffer address\n"); + return -1; + } + + memcpy((void *)hdr_addr, &hdr, sizeof(struct virtio_net_hdr)); + if (offset >= data_len) { + hdr_desc->flags &= ~VIRTIO_NEXT_FLAG; + } + + for (buf_idx = hdr_desc->next; offset < data_len; buf_idx = q->desc[hdr_idx].next) { + 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); + offset += len; + if (offset < data_len) { + buf_desc->flags = VIRTIO_NEXT_FLAG; + } + 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; + q->cur_avail_idx++; + if (q->cur_avail_idx < last_idx) + q->idx_overflow = false; } else { - *status = NIC_STATUS_OK; + 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 } - PrintDebug("Returning Status: %d\n", *status); + if (!(q->avail->flags & VIRTIO_NO_IRQ_FLAG)) { + PrintDebug("Raising IRQ %d\n", virtio->pci_dev->config_header.intr_line); + v3_pci_raise_irq(virtio->virtio_dev->pci_bus, 0, virtio->pci_dev); + virtio->virtio_cfg.pci_isr = 0x1; + } - return 0; +#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; } +// TODO: +int virtio_send(struct guest_info * vm, char *dev_name, uchar_t * buf, uint_t size) { + struct vm_device * virtio_dev = v3_find_dev(vm, dev_name); -//get packet from network, and send to guest -static int handle_pkt_read(struct vm_device *dev, struct virtio_net_hdr *hdr, - struct vring_desc *buf_desc, uint8_t *status) -{ - //struct virtio_net_state * virtio = (struct virtio_net_state *)dev->private_data; - uint8_t * buf = NULL; + // TODO: how to get virtio net state from device?? + 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); +} - PrintDebug("Handling Virtio Net read\n"); - if (guest_pa_to_host_va(dev->vm, buf_desc->addr_gpa, (addr_t *)&(buf)) == -1) { - PrintError("Could not translate buffer address\n"); - return -1; +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); - PrintDebug("Length=%d\n", buf_desc->length); + return pkt->size; +} - if (read_op(dev, buf, buf_desc->length) == -1) { - *status = NIC_STATUS_ERR; - return -1; - } else { - *status = NIC_STATUS_OK; - } - PrintDebug("Returning Status: %d\n", *status); +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 __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) -{ + +static int get_desc_count(struct virtio_queue * q, int index) { struct vring_desc * tmp_desc = &(q->desc[index]); int cnt = 1; @@ -254,249 +405,189 @@ static int get_desc_count(struct virtio_queue * q, int index) } -static int handle_ctrl(struct vm_device * dev) { - - - return 0; -} - -// TODO: handle receiving, not done yet -//send packet to guest -static int handle_pkt_rx(struct vm_device * dev) -{ - - if (handle_pkt_read(dev, NULL, 0, NULL) == -1) { - PrintError("Error handling nic operation\n"); - return -1; - } - +static int handle_ctrl(struct virtio_net_state * dev) { return 0; } //get packet from guest -static int handle_pkt_tx(struct vm_device * dev) +static int handle_pkt_tx(struct virtio_net_state * virtio_state) { - struct virtio_net_state *virtio = (struct virtio_net_state *)dev->private_data; - struct virtio_queue *q = &(virtio->rx_vq); + struct virtio_queue * q = &(virtio_state->tx_vq); + struct virtio_net_hdr * hdr = NULL; - PrintDebug("VIRTIO NIC KICK: cur_index=%d (mod=%d), avail_index=%d\n", - q->cur_avail_idx, q->cur_avail_idx % QUEUE_SIZE, q->avail->index); + if (q->avail->index < q->last_avail_idx) + q->idx_overflow = true; + q->last_avail_idx = q->avail->index; - while (q->cur_avail_idx < q->avail->index) { + 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; - struct vring_desc * buf_desc = NULL; - struct vring_desc * status_desc = NULL; - struct virtio_net_hdr hdr; addr_t hdr_addr = 0; - uint16_t desc_idx = q->avail->ring[q->cur_avail_idx % QUEUE_SIZE]; + uint16_t desc_idx = q->avail->ring[q->cur_avail_idx % q->queue_size]; int desc_cnt = get_desc_count(q, desc_idx); - int i = 0; - uint8_t * status_ptr = NULL; - uint8_t status = NIC_STATUS_OK; uint32_t req_len = 0; - - PrintDebug("Descriptor Count=%d, index=%d\n", desc_cnt, q->cur_avail_idx % QUEUE_SIZE); + int i = 0; hdr_desc = &(q->desc[desc_idx]); - - PrintDebug("Header Descriptor (ptr=%p) gpa=%p, len=%d, flags=%x, next=%d\n", hdr_desc, - (void *)(hdr_desc->addr_gpa), hdr_desc->length, hdr_desc->flags, hdr_desc->next); - - if (guest_pa_to_host_va(dev->vm, hdr_desc->addr_gpa, &(hdr_addr)) == -1) { + if (guest_pa_to_host_va(virtio_state->virtio_dev->vm, hdr_desc->addr_gpa, &(hdr_addr)) == -1) { PrintError("Could not translate block header address\n"); return -1; } - // We copy the block op header out because we are going to modify its contents - memcpy(&hdr, (void *)hdr_addr, sizeof(struct virtio_net_hdr)); - - PrintDebug("NIC Op Hdr (ptr=%p) type=%d, sector=%p\n", (void *)hdr_addr, hdr.hdr_len, (void *)hdr.csum_start); - + hdr = (struct virtio_net_hdr*)hdr_addr; desc_idx = hdr_desc->next; - - for (i = 0; i < desc_cnt - 2; i++) { - uint8_t tmp_status = NIC_STATUS_OK; - - buf_desc = &(q->desc[desc_idx]); - - PrintDebug("Buffer Descriptor (ptr=%p) gpa=%p, len=%d, flags=%x, next=%d\n", buf_desc, - (void *)(buf_desc->addr_gpa), buf_desc->length, buf_desc->flags, buf_desc->next); - - if (handle_pkt_write(dev, &hdr, buf_desc, &tmp_status) == -1) { + + for (i = 0; i < desc_cnt - 1; i++) { + struct vring_desc * buf_desc = &(q->desc[desc_idx]); + if (pkt_write(virtio_state, buf_desc) == -1) { PrintError("Error handling nic operation\n"); return -1; } - if (tmp_status != NIC_STATUS_OK) { - status = tmp_status; - } - req_len += buf_desc->length; desc_idx = buf_desc->next; } + virtio_state->pkt_sent ++; - status_desc = &(q->desc[desc_idx]); - - PrintDebug("Status Descriptor (ptr=%p) gpa=%p, len=%d, flags=%x, next=%d\n", status_desc, - (void *)(status_desc->addr_gpa), status_desc->length, status_desc->flags, status_desc->next); - - if (guest_pa_to_host_va(dev->vm, status_desc->addr_gpa, (addr_t *)&(status_ptr)) == -1) { - PrintError("Could not translate status address\n"); - return -1; - } - - req_len += status_desc->length; - *status_ptr = status; - - q->used->ring[q->used->index % QUEUE_SIZE].id = q->avail->ring[q->cur_avail_idx % QUEUE_SIZE]; - q->used->ring[q->used->index % QUEUE_SIZE].length = req_len; // What do we set this to???? - + 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++; - q->cur_avail_idx++; + + int last_idx = q->cur_avail_idx; + q->cur_avail_idx ++; + if (q->cur_avail_idx < last_idx) + q->idx_overflow = false; } if (!(q->avail->flags & VIRTIO_NO_IRQ_FLAG)) { - PrintDebug("Raising IRQ %d\n", virtio->pci_dev->config_header.intr_line); - v3_pci_raise_irq(virtio->pci_bus, 0, virtio->pci_dev); - virtio->virtio_cfg.pci_isr = 0x1; + v3_pci_raise_irq(virtio_state->virtio_dev->pci_bus, 0, virtio_state->pci_dev); + 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 vm_device * dev, struct virtio_queue *queue, addr_t pfn, addr_t page_addr) -{ +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; - queue->ring_desc_addr = page_addr ; - queue->ring_avail_addr = page_addr + (QUEUE_SIZE * sizeof(struct vring_desc)); - queue->ring_used_addr = (queue->ring_avail_addr + \ - sizeof(struct vring_avail) + \ - (QUEUE_SIZE * sizeof(uint16_t))); + queue->ring_desc_addr = page_addr; + queue->ring_avail_addr = page_addr + (queue->queue_size * sizeof(struct vring_desc)); + queue->ring_used_addr = ((queue->ring_avail_addr) + + (sizeof(struct vring_avail)) + + (queue->queue_size * sizeof(uint16_t))); // round up to next page boundary. queue->ring_used_addr = (queue->ring_used_addr + 0xfff) & ~0xfff; - - if (guest_pa_to_host_va(dev->vm, queue->ring_desc_addr, (addr_t *)&(queue->desc)) == -1) { + 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(dev->vm, queue->ring_avail_addr, (addr_t *)&(queue->avail)) == -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(dev->vm, queue->ring_used_addr, (addr_t *)&(queue->used)) == -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; } PrintDebug("RingDesc_addr=%p, Avail_addr=%p, Used_addr=%p\n", - (void *)(queue->ring_desc_addr), - (void *)(queue->ring_avail_addr), - (void *)(queue->ring_used_addr)); - + (void *)(queue->ring_desc_addr), + (void *)(queue->ring_avail_addr), + (void *)(queue->ring_used_addr)); + PrintDebug("RingDesc=%p, Avail=%p, Used=%p\n", - queue->desc, queue->avail, queue->used); - + queue->desc, queue->avail, queue->used); + 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 *)dev->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 Write for port %d (index=%d) len=%d, value=%x\n", + 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) { - 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(dev, &virtio->rx_vq, pfn, page_addr); - break; - case 1: - virtio_setup_queue(dev, &virtio->tx_vq, pfn, page_addr); - break; - case 2: - virtio_setup_queue(dev, &virtio->ctrl_vq, pfn, page_addr); - break; - - default: - break; - } - } else { + 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); + break; + case 1: + virtio_setup_queue(virtio, &virtio->tx_vq, pfn, page_addr); + break; + case 2: + virtio_setup_queue(virtio, &virtio->ctrl_vq, pfn, page_addr); + 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: - PrintDebug("Handling Kick\n"); - uint16_t queue_idx = *(uint16_t *)src; - if (queue_idx == 0){ - if (handle_pkt_rx(dev) == -1) { - PrintError("Could not handle NIC Notification\n"); - return -1; - } - }else if (queue_idx == 1){ - if (handle_pkt_tx(dev) == -1) { + + case VRING_Q_NOTIFY_PORT: + { + 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){ + if (handle_pkt_tx(virtio) == -1) { PrintError("Could not handle NIC Notification\n"); return -1; } - }else if (queue_idx == 2){ - if (handle_ctrl(dev) == -1) { + } else if (queue_idx == 2){ + if (handle_ctrl(virtio) == -1) { PrintError("Could not handle NIC Notification\n"); return -1; } - }else { - PrintError("Virtio NIC device only uses 3 queue, selected %d\n", - queue_idx); + } else { + PrintError("Virtio NIC device only uses 3 queue, selected %d\n", + queue_idx); + } + break; } - - break; + case VIRTIO_STATUS_PORT: virtio->virtio_cfg.status = *(uint8_t *)src; - if (virtio->virtio_cfg.status == 0) { PrintDebug("Resetting device\n"); - virtio_reset(dev); + virtio_init_state(virtio); } - break; case VIRTIO_ISR_PORT: @@ -510,89 +601,87 @@ 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; - struct virtio_net_state * virtio = (struct virtio_net_state *)dev->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; uint16_t queue_idx = virtio->virtio_cfg.vring_queue_selector; - - PrintDebug("VIRTIO NIC Read for port %d (index =%d), length=%d\n", + PrintDebug("Virtio NIC %p: Read for port %d (index =%d), length=%d", private_data, port, port_idx, length); - + switch (port_idx) { case HOST_FEATURES_PORT: if (length != 4) { 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; - break; - case 1: - *(uint32_t *)dst = virtio->tx_vq.pfn; - break; - case 2: - *(uint32_t *)dst = virtio->ctrl_vq.pfn; - break; - default: - break; - } - + *(uint32_t *)dst = virtio->rx_vq.pfn; + break; + case 1: + *(uint32_t *)dst = virtio->tx_vq.pfn; + break; + case 2: + *(uint32_t *)dst = virtio->ctrl_vq.pfn; + break; + default: + break; + } + PrintDebug(", value=0x%x\n", *(uint32_t *)dst); break; + 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; - break; - case 1: - *(uint32_t *)dst = virtio->tx_vq.queue_size; - break; - case 2: - *(uint32_t *)dst = virtio->ctrl_vq.queue_size; - break; - default: - break; - } - + *(uint16_t *)dst = virtio->rx_vq.queue_size; + break; + case 1: + *(uint16_t *)dst = virtio->tx_vq.queue_size; + break; + case 2: + *(uint16_t *)dst = virtio->ctrl_vq.queue_size; + break; + default: + break; + } + PrintDebug("queue index: %d, value=0x%x\n", (int)queue_idx, *(uint16_t *)dst); break; + 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->pci_bus, 0, virtio->pci_dev); + v3_pci_lower_irq(virtio->virtio_dev->pci_bus, 0, virtio->pci_dev); + PrintDebug(", value=0x%x\n", *(uint8_t *)dst); break; default: - PrintError("Read of Unhandled Virtio Read\n"); - return -1; + PrintError("Virtio NIC: Read of Unhandled Virtio Read\n"); + return -1; } return length; @@ -601,124 +690,228 @@ static int virtio_io_read(uint16_t port, void * dst, uint_t length, void * priva static struct v3_device_ops dev_ops = { .free = virtio_free, - .reset = virtio_reset, + .reset = NULL, .start = NULL, .stop = NULL, }; +static int register_dev(struct virtio_dev_state * virtio, struct virtio_net_state * net_state) +{ + struct pci_device * pci_dev = NULL; + struct v3_pci_bar bars[6]; + int num_ports = sizeof(struct virtio_config); + int tmp_ports = num_ports; + int i; + + // 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; + } + + // this is to account for any low order bits being set in num_ports + // if there are none, then num_ports was already a power of 2 so we shift right to reset it + if ((num_ports & ((net_state->io_range_size >> 1) - 1)) == 0) { + net_state->io_range_size >>= 1; + } + + for (i = 0; i < 6; i++) { + bars[i].type = PCI_BAR_NONE; + } + + PrintDebug("Virtio-NIC io_range_size = %d\n", net_state->io_range_size); + + 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; + + pci_dev = v3_pci_register_device(virtio->pci_bus, PCI_STD_DEVICE, + 0, PCI_AUTO_DEV_NUM, 0, + "LNX_VIRTIO_NIC", bars, + NULL, NULL, NULL, net_state); + + if (!pci_dev) { + PrintError("Virtio NIC: Could not register PCI Device\n"); + return -1; + } -int v3_virtio_register_nic(struct vm_device *dev, struct v3_net_ops *ops, void *private_data) { - struct virtio_net_state * virtio = (struct virtio_net_state *)dev->private_data; + PrintDebug("Virtio NIC: registered to PCI bus\n"); - virtio->net_ops = ops; + pci_dev->config_header.vendor_id = VIRTIO_VENDOR_ID; + pci_dev->config_header.subsystem_vendor_id = VIRTIO_SUBVENDOR_ID; + + + 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.intr_pin = 1; + pci_dev->config_header.max_latency = 1; // ?? (qemu does it...) + + 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; + + virtio_init_state(net_state); return 0; } +static int connect_fn(struct guest_info * info, + void * frontend_data, + struct v3_dev_net_ops * ops, + v3_cfg_tree_t * cfg, + void * private_data) { + struct virtio_dev_state * virtio = (struct virtio_dev_state *)frontend_data; + struct virtio_net_state * net_state = (struct virtio_net_state *)V3_Malloc(sizeof(struct virtio_net_state)); -static int virtio_init(struct guest_info * vm, void *cfg_data) { - struct vm_device * pci_bus = v3_find_dev(vm, (char *)cfg_data); - struct virtio_net_state * virtio_state = NULL; - struct pci_device * pci_dev = NULL; + memset(net_state, 0, sizeof(struct virtio_net_state)); - PrintDebug("Initializing VIRTIO Network device\n"); + register_dev(virtio, net_state); - if (pci_bus == NULL) { - PrintError("VirtIO network devices require a PCI Bus"); - return -1; - } - - virtio_state = (struct virtio_net_state *)V3_Malloc(sizeof(struct virtio_net_state)); - memset(virtio_state, 0, sizeof(struct virtio_net_state)); + net_state->net_ops = ops; + net_state->backend_data = private_data; - struct vm_device * dev = v3_allocate_device("LNX_VIRTIO_NIC", &dev_ops, virtio_state); - if (v3_attach_device(vm, dev) == -1) { - PrintError("Could not attach device %s\n", "LNX_VIRTIO_NIC"); - return -1; - } + return 0; +} +#if 1 +//temporary interface between Virtio-NIC and Vnet +//Treat vnet as backend, and virtio nic as frontend - // PCI initialization - { - struct v3_pci_bar bars[6]; - int num_ports = sizeof(struct virtio_config); - int tmp_ports = num_ports; - int i; +//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); - // This gets the number of ports, rounded up to a power of 2 - virtio_state->io_range_size = 1; // must be a power of 2 + v3_vnet_send_rawpkt(buf, len, private_data); + return 0; +} - while (tmp_ports > 0) { - tmp_ports >>= 1; - virtio_state->io_range_size <<= 1; - } - - // this is to account for any low order bits being set in num_ports - // if there are none, then num_ports was already a power of 2 so we shift right to reset it - if ((num_ports & ((virtio_state->io_range_size >> 1) - 1)) == 0) { - virtio_state->io_range_size >>= 1; - } +//used to send packet to guest by a virtio nic +static int vnet_receive(uint8_t * buf, uint32_t count, void * private_data, struct vm_device *src_dev){ - for (i = 0; i < 6; i++) { - bars[i].type = PCI_BAR_NONE; - } + return 0; +} - PrintDebug("Virtio-NIC io_range_size = %d\n", virtio_state->io_range_size); +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); - bars[0].type = PCI_BAR_IO; - bars[0].default_base_port = -1; - bars[0].num_ports = virtio_state->io_range_size; + return __virtio_dev_send(buf, len, private_data); +} - bars[0].io_read = virtio_io_read; - bars[0].io_write = virtio_io_write; - bars[0].private_data = dev; - pci_dev = v3_pci_register_device(pci_bus, PCI_STD_DEVICE, - 0, PCI_AUTO_DEV_NUM, 0, - "LNX_VIRTIO_NIC", bars, - NULL, NULL, NULL, dev, NULL); +//register a virtio device to the vnet as backend +void register_virtio_to_vnet(struct vm_device *dev, + char *dev_name, + uchar_t mac[6]){ + 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}; - if (!pci_dev) { - PrintError("Could not register PCI Device\n"); - return -1; - } - - pci_dev->config_header.vendor_id = VIRTIO_VENDOR_ID; - pci_dev->config_header.subsystem_vendor_id = VIRTIO_SUBVENDOR_ID; + + 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 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); + vnet_add_route_entry(zeromac, dstmac, MAC_ANY, MAC_NONE, 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; +} - 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; +#endif - // TODO:how to define new one for virtio net device - pci_dev->config_header.subsystem_id = VIRTIO_BLOCK_SUBDEVICE_ID; +static int virtio_init(struct guest_info * vm, v3_cfg_tree_t * cfg) { + struct vm_device * pci_bus = v3_find_dev(vm, v3_cfg_val(cfg, "bus")); + struct virtio_dev_state * virtio_state = NULL; + char * name = v3_cfg_val(cfg, "name"); + PrintDebug("Virtio NIC: Initializing VIRTIO Network device: %s\n", name); - pci_dev->config_header.intr_pin = 1; + if (pci_bus == NULL) { + PrintError("Virtio NIC: VirtIO devices require a PCI Bus"); + return -1; + } - pci_dev->config_header.max_latency = 1; // ?? (qemu does it...) + virtio_state = (struct virtio_dev_state *)V3_Malloc(sizeof(struct virtio_dev_state)); + memset(virtio_state, 0, sizeof(struct virtio_dev_state)); + INIT_LIST_HEAD(&(virtio_state->dev_list)); + virtio_state->pci_bus = pci_bus; + virtio_state->vm = vm; - virtio_state->pci_dev = pci_dev; - virtio_state->pci_bus = pci_bus; + struct vm_device * dev = v3_allocate_device(name, &dev_ops, virtio_state); + if (v3_attach_device(vm, dev) == -1) { + PrintError("Virtio NIC: Could not attach device %s\n", name); + return -1; } - virtio_state->virtio_cfg.host_features = 0; //no features support now + if (v3_dev_add_net_frontend(vm, name, connect_fn, (void *)virtio_state) == -1) { + PrintError("Virtio NIC: Could not register %s as net frontend\n", name); + return -1; + } - virtio_state->rx_vq.queue_size = QUEUE_SIZE; - virtio_state->tx_vq.queue_size = QUEUE_SIZE; - virtio_state->ctrl_vq.queue_size = CTRL_QUEUE_SIZE; - + //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)); + 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 = &send; + net_state->net_ops->receive = &receive; + + 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); - virtio_reset(dev); + temp_net_states[net_idx ++] = net_state; -// TODO: net ops - virtio_state->net_ops = NULL; + 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); + + 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; } - device_register("LNX_VIRTIO_NIC", virtio_init) - +