X-Git-Url: http://v3vee.org/palacios/gitweb/gitweb.cgi?a=blobdiff_plain;f=palacios%2Fsrc%2Fdevices%2Flnx_virtio_nic.c;h=da6644ea753672fbbefcdd62b966c25f0a3afbcf;hb=322b52c9b40b5d043e064efc75546804573a3be3;hp=a9825b96677cc1aed27d4806b4926e78123d33ee;hpb=00460ac661afa22c2bb495eba4d4655d20a1bfdc;p=palacios.git diff --git a/palacios/src/devices/lnx_virtio_nic.c b/palacios/src/devices/lnx_virtio_nic.c index a9825b9..da6644e 100644 --- a/palacios/src/devices/lnx_virtio_nic.c +++ b/palacios/src/devices/lnx_virtio_nic.c @@ -29,6 +29,7 @@ #include #include #include +#include #ifndef CONFIG_DEBUG_VIRTIO_NET @@ -36,13 +37,12 @@ #define PrintDebug(fmt, args...) #endif -#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 { uint8_t flags; - -#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 */ @@ -51,8 +51,6 @@ struct virtio_net_hdr { }__attribute__((packed)); -/* This is the version of the header to use when the MRG_RXBUF - * feature has been negotiated. */ struct virtio_net_hdr_mrg_rxbuf { struct virtio_net_hdr hdr; uint16_t num_buffers; /* Number of merged rx buffers */ @@ -62,13 +60,15 @@ struct virtio_net_hdr_mrg_rxbuf { #define TX_QUEUE_SIZE 64 #define RX_QUEUE_SIZE 1024 #define CTRL_QUEUE_SIZE 64 -#define ETH_ALEN 6 #define VIRTIO_NET_F_MRG_RXBUF 15 /* Host can merge receive buffers. */ #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_HOST_TSO4 11 /* Host can handle TSOv4 in. */ +/* Port to get virtio config */ +#define VIRTIO_NET_CONFIG 20 + struct virtio_net_config { uint8_t mac[ETH_ALEN]; /* VIRTIO_NET_F_MAC */ @@ -79,6 +79,8 @@ struct virtio_dev_state { struct vm_device * pci_bus; struct list_head dev_list; struct v3_vm_info *vm; + + uint8_t mac[ETH_ALEN]; }; struct virtio_net_state { @@ -89,14 +91,11 @@ struct virtio_net_state { struct pci_device * pci_dev; int io_range_size; - struct virtio_queue rx_vq; /* idx 0, pkts to guest */ - struct virtio_queue tx_vq; /* idx 1, pkts from guest */ - struct virtio_queue ctrl_vq; /* idx 2 */ - - int buffed_rx; - int tx_disabled; /* stop TX pkts from guest */ + struct virtio_queue rx_vq; /* idx 0*/ + struct virtio_queue tx_vq; /* idx 1*/ + struct virtio_queue ctrl_vq; /* idx 2*/ - uint64_t pkt_sent, pkt_recv, pkt_drop; + struct nic_statistics statistics; struct v3_dev_net_ops * net_ops; v3_lock_t rx_lock, tx_lock; @@ -106,20 +105,6 @@ struct virtio_net_state { struct list_head dev_link; }; -/* virtio nic error type */ -#define ERR_VIRTIO_OTHER 1 -#define ERR_VIRTIO_RXQ_FULL 2 -#define ERR_VIRTIO_RXQ_NOSET 3 -#define ERR_VIRTIO_TXQ_NOSET 4 -#define ERR_VIRTIO_TXQ_FULL 5 -#define ERR_VIRTIO_TXQ_DISABLED 6 - - -static int virtio_free(struct vm_device * dev) -{ - - return 0; -} static int virtio_init_state(struct virtio_net_state * virtio) { @@ -147,21 +132,17 @@ static int virtio_init_state(struct virtio_net_state * virtio) virtio->virtio_cfg.pci_isr = 0; - virtio->virtio_cfg.host_features = 0; // (1 << VIRTIO_NET_F_MAC); + virtio->virtio_cfg.host_features = 0 | (1 << VIRTIO_NET_F_MAC); if ((v3_lock_init(&(virtio->rx_lock)) == -1) || (v3_lock_init(&(virtio->tx_lock)) == -1)){ PrintError("Virtio NIC: Failure to init locks for net_state\n"); } - virtio->pkt_sent = virtio->pkt_recv = virtio->pkt_drop = 0; - virtio->buffed_rx = 0; - return 0; } -static int -pkt_tx(struct guest_info * core, +static int tx_one_pkt(struct guest_info * core, struct virtio_net_state * virtio, struct vring_desc * buf_desc) { @@ -169,11 +150,20 @@ pkt_tx(struct guest_info * core, uint32_t len = buf_desc->length; if (v3_gpa_to_hva(core, buf_desc->addr_gpa, (addr_t *)&(buf)) == -1) { - PrintError("Could not translate buffer address\n"); - return -ERR_VIRTIO_OTHER; + PrintDebug("Could not translate buffer address\n"); + return -1; } - return virtio->net_ops->send(buf, len, virtio->backend_data, NULL); + if(virtio->net_ops->send(buf, len, virtio->backend_data) >= 0){ + virtio->statistics.tx_pkts ++; + virtio->statistics.tx_bytes += len; + + return 0; + } + + virtio->statistics.tx_dropped ++; + + return -1; } @@ -220,50 +210,37 @@ static inline void disable_cb(struct virtio_queue *queue) { } -/* interrupt the guest, so the guest core get EXIT to Palacios - * this happens when there are either incoming pkts for the guest - * or the guest can start TX pkts again */ +/* interrupt the guest, so the guest core get EXIT to Palacios */ static inline void notify_guest(struct virtio_net_state * virtio){ v3_interrupt_cpu(virtio->virtio_dev->vm, virtio->virtio_dev->vm->cores[0].cpu_id, 0); } -/* guest free some pkts from rx queue */ -static int handle_rx_kick(struct guest_info *core, +/* guest free some pkts for rx queue */ +static int handle_rx_queue_kick(struct guest_info * core, struct virtio_net_state * virtio) { - unsigned long flags; - - flags = v3_lock_irqsave(virtio->rx_lock); - - virtio->net_ops->start_rx(virtio->backend_data); - //disable_cb(&virtio->rx_vq); - - v3_unlock_irqrestore(virtio->rx_lock, flags); - return 0; } -static int handle_ctrl(struct guest_info *core, +static int handle_ctrl(struct guest_info * core, struct virtio_net_state * virtio) { return 0; } -static int handle_pkt_tx(struct guest_info *core, +static int handle_pkt_tx(struct guest_info * core, struct virtio_net_state * virtio_state) { - struct virtio_queue * q = &(virtio_state->tx_vq); - struct virtio_net_hdr * hdr = NULL; - int recved = 0; + struct virtio_queue *q = &(virtio_state->tx_vq); + struct virtio_net_hdr *hdr = NULL; + int txed = 0; unsigned long flags; - if (!q->ring_avail_addr) - return -ERR_VIRTIO_TXQ_NOSET; - - if(virtio_state->tx_disabled) - return -ERR_VIRTIO_TXQ_DISABLED; + if (!q->ring_avail_addr) { + return -1; + } flags = v3_lock_irqsave(virtio_state->tx_lock); while (q->cur_avail_idx != q->avail->index) { @@ -280,7 +257,7 @@ static int handle_pkt_tx(struct guest_info *core, goto exit_error; } - hdr = (struct virtio_net_hdr*)hdr_addr; + hdr = (struct virtio_net_hdr *)hdr_addr; desc_idx = hdr_desc->next; if(desc_cnt > 2){ @@ -291,7 +268,7 @@ static int handle_pkt_tx(struct guest_info *core, /* here we assumed that one ethernet pkt is not splitted into multiple virtio buffer */ for (i = 0; i < desc_cnt - 1; i++) { struct vring_desc * buf_desc = &(q->desc[desc_idx]); - if (pkt_tx(core, virtio_state, buf_desc) == -1) { + if (tx_one_pkt(core, virtio_state, buf_desc) == -1) { PrintError("Error handling nic operation\n"); goto exit_error; } @@ -299,24 +276,23 @@ static int handle_pkt_tx(struct guest_info *core, req_len += buf_desc->length; desc_idx = buf_desc->next; } - virtio_state->pkt_sent ++; - recved ++; 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->ring[q->used->index % q->queue_size].length = req_len; /* What do we set this to???? */ q->used->index ++; q->cur_avail_idx ++; + + txed ++; } v3_unlock_irqrestore(virtio_state->tx_lock, flags); - - if(!recved) - return 0; - if (!(q->avail->flags & VIRTIO_NO_IRQ_FLAG)) { + if (txed && !(q->avail->flags & VIRTIO_NO_IRQ_FLAG)) { v3_pci_raise_irq(virtio_state->virtio_dev->pci_bus, 0, virtio_state->pci_dev); virtio_state->virtio_cfg.pci_isr = 0x1; + + virtio_state->statistics.interrupts ++; } return 0; @@ -324,7 +300,7 @@ static int handle_pkt_tx(struct guest_info *core, exit_error: v3_unlock_irqrestore(virtio_state->tx_lock, flags); - return -ERR_VIRTIO_OTHER; + return -1; } @@ -398,11 +374,9 @@ static int virtio_io_write(struct guest_info *core, switch (queue_idx) { case 0: virtio_setup_queue(core, virtio, &virtio->rx_vq, pfn, page_addr); - //disable_cb(&virtio->rx_vq); break; case 1: virtio_setup_queue(core, virtio, &virtio->tx_vq, pfn, page_addr); - //disable_cb(&virtio->tx_vq); break; case 2: virtio_setup_queue(core, virtio, &virtio->ctrl_vq, pfn, page_addr); @@ -425,15 +399,18 @@ static int virtio_io_write(struct guest_info *core, { uint16_t queue_idx = *(uint16_t *)src; if (queue_idx == 0){ - handle_rx_kick(core, virtio); + if(handle_rx_queue_kick(core, virtio) == -1){ + PrintError("Could not handle Virtio NIC rx kick\n"); + return -1; + } } else if (queue_idx == 1){ if (handle_pkt_tx(core, virtio) == -1) { - PrintError("Could not handle NIC Notification\n"); + PrintError("Could not handle Virtio NIC tx kick\n"); return -1; } } else if (queue_idx == 2){ if (handle_ctrl(core, virtio) == -1) { - PrintError("Could not handle NIC Notification\n"); + PrintError("Could not handle Virtio NIC ctrl kick\n"); return -1; } } else { @@ -445,7 +422,6 @@ static int virtio_io_write(struct guest_info *core, case VIRTIO_STATUS_PORT: virtio->virtio_cfg.status = *(uint8_t *)src; if (virtio->virtio_cfg.status == 0) { - PrintDebug("Resetting device\n"); virtio_init_state(virtio); } break; @@ -470,7 +446,7 @@ static int virtio_io_read(struct guest_info *core, int port_idx = port % virtio->io_range_size; uint16_t queue_idx = virtio->virtio_cfg.vring_queue_selector; - PrintDebug("Virtio NIC %p: Read for port %d (index =%d), length=%d\n", private_data, + PrintDebug("Virtio NIC %p: Read for port 0x%x (index =%d), length=%d\n", private_data, port, port_idx, length); switch (port_idx) { @@ -536,6 +512,10 @@ static int virtio_io_read(struct guest_info *core, v3_pci_lower_irq(virtio->virtio_dev->pci_bus, 0, virtio->pci_dev); break; + case VIRTIO_NET_CONFIG ... VIRTIO_NET_CONFIG + ETH_ALEN: + *(uint8_t *)dst = virtio->net_cfg.mac[port_idx-VIRTIO_NET_CONFIG]; + break; + default: PrintError("Virtio NIC: Read of Unhandled Virtio Read:%d\n", port_idx); return -1; @@ -545,39 +525,29 @@ static int virtio_io_read(struct guest_info *core, } +/* receiving raw ethernet pkt from backend */ static int virtio_rx(uint8_t * buf, uint32_t size, void * private_data) { struct virtio_net_state * virtio = (struct virtio_net_state *)private_data; struct virtio_queue * q = &(virtio->rx_vq); struct virtio_net_hdr_mrg_rxbuf hdr; uint32_t hdr_len = sizeof(struct virtio_net_hdr_mrg_rxbuf); - uint32_t data_len = size; + uint32_t data_len; uint32_t offset = 0; unsigned long flags; - int ret_val = -ERR_VIRTIO_OTHER; - int raw = 1; #ifndef CONFIG_DEBUG_VIRTIO_NET - { - PrintDebug("Virtio-NIC: virtio_rx: size: %d\n", size); - //v3_hexdump(buf, size, NULL, 0); - } + PrintDebug("Virtio-NIC: virtio_rx: size: %d\n", size); + v3_hexdump(buf, size, NULL, 0); #endif flags = v3_lock_irqsave(virtio->rx_lock); - virtio->pkt_recv ++; - if (!raw) - data_len -= hdr_len; - - if (!raw) - memcpy(&hdr, buf, sizeof(struct virtio_net_hdr_mrg_rxbuf)); - else - memset(&hdr, 0, sizeof(struct virtio_net_hdr_mrg_rxbuf)); + data_len = size; + memset(&hdr, 0, sizeof(struct virtio_net_hdr_mrg_rxbuf)); if (q->ring_avail_addr == 0) { - PrintError("Queue is not set\n"); - ret_val = -ERR_VIRTIO_RXQ_NOSET; - goto exit; + PrintDebug("Queue is not set\n"); + goto err_exit; } if (q->cur_avail_idx != q->avail->index){ @@ -588,8 +558,8 @@ static int virtio_rx(uint8_t * buf, uint32_t size, void * private_data) { hdr_desc = &(q->desc[hdr_idx]); if (v3_gpa_to_hva(&(virtio->virtio_dev->vm->cores[0]), hdr_desc->addr_gpa, &(hdr_addr)) == -1) { - PrintError("Could not translate receive buffer address\n"); - goto exit; + PrintDebug("Could not translate receive buffer address\n"); + goto err_exit; } hdr.num_buffers = 1; memcpy((void *)hdr_addr, &hdr, sizeof(struct virtio_net_hdr_mrg_rxbuf)); @@ -616,88 +586,61 @@ static int virtio_rx(uint8_t * buf, uint32_t size, void * private_data) { q->used->index++; q->cur_avail_idx++; - /* if there are certain num of pkts in the RX queue, notify guest - * so guest will exit to palacios - * when it returns, guest gets the virtio rx interrupt */ - if((++virtio->buffed_rx > q->queue_size/5) && - (q->avail->flags & VIRTIO_NO_IRQ_FLAG)) { - if(virtio->virtio_dev->vm->cores[0].cpu_id != V3_Get_CPU()){ - notify_guest(virtio); - } - virtio->buffed_rx = 0; - } + virtio->statistics.rx_pkts ++; + virtio->statistics.rx_bytes += size; } else { - virtio->pkt_drop++; - /* RX queue is full, tell backend to stop RX on this device */ - virtio->net_ops->stop_rx(virtio->backend_data); - enable_cb(&virtio->rx_vq); + virtio->statistics.rx_dropped ++; - ret_val = -ERR_VIRTIO_RXQ_FULL; - goto exit; + goto err_exit; } 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; + virtio->statistics.interrupts ++; } - ret_val = offset; + v3_unlock_irqrestore(virtio->rx_lock, flags); + + return 0; -exit: +err_exit: v3_unlock_irqrestore(virtio->rx_lock, flags); - return ret_val; + return -1; } +static int virtio_free(struct virtio_dev_state * virtio) { + struct virtio_net_state * backend = NULL; + struct virtio_net_state * tmp = NULL; -static struct v3_device_ops dev_ops = { - .free = virtio_free, - .reset = NULL, - .start = NULL, - .stop = NULL, -}; - - -/* TODO: Issue here: which vm info it needs? calling VM or the device's own VM? */ -static void virtio_nic_poll(struct v3_vm_info * vm, void * data){ - struct virtio_net_state * virtio = (struct virtio_net_state *)data; - - handle_pkt_tx(&(vm->cores[0]), virtio); -} -static void virtio_start_tx(void * data){ - struct virtio_net_state * virtio = (struct virtio_net_state *)data; - unsigned long flags; + list_for_each_entry_safe(backend, tmp, &(virtio->dev_list), dev_link) { - flags = v3_lock_irqsave(virtio->tx_lock); - virtio->tx_disabled = 0; + // unregister from PCI - /* notify the device's guest to start sending pkt */ - if(virtio->virtio_dev->vm->cores[0].cpu_id != V3_Get_CPU()){ - notify_guest(virtio); + list_del(&(backend->dev_link)); + V3_Free(backend); } - v3_unlock_irqrestore(virtio->tx_lock, flags); -} -static void virtio_stop_tx(void * data){ - struct virtio_net_state * virtio = (struct virtio_net_state *)data; - unsigned long flags; + V3_Free(virtio); + return 0; +} - flags = v3_lock_irqsave(virtio->tx_lock); - virtio->tx_disabled = 1; - /* stop the guest to exit to palacios for sending pkt? */ - if(virtio->virtio_dev->vm->cores[0].cpu_id != V3_Get_CPU()){ - disable_cb(&virtio->tx_vq); - } +static struct v3_device_ops dev_ops = { + .free = (int (*)(void *))virtio_free, +}; - v3_unlock_irqrestore(virtio->tx_lock, flags); -} +static void virtio_nic_poll(struct v3_vm_info * vm, void * data){ + struct virtio_net_state * virtio = (struct virtio_net_state *)data; - + handle_pkt_tx(&(vm->cores[0]), virtio); +} static int register_dev(struct virtio_dev_state * virtio, struct virtio_net_state * net_state) @@ -715,8 +658,9 @@ static int register_dev(struct virtio_dev_state * virtio, 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 + /* 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; } @@ -760,13 +704,13 @@ static int register_dev(struct virtio_dev_state * virtio, net_state->pci_dev = pci_dev; net_state->virtio_dev = virtio; - uchar_t mac[6] = {0x11,0x11,0x11,0x11,0x11,0x11}; - memcpy(net_state->net_cfg.mac, mac, 6); - - memcpy(pci_dev->config_data, net_state->net_cfg.mac, ETH_ALEN); - + memcpy(net_state->net_cfg.mac, virtio->mac, 6); + virtio_init_state(net_state); + /* Add backend to list of devices */ + list_add(&(net_state->dev_link), &(virtio->dev_list)); + return 0; } @@ -783,12 +727,13 @@ static int connect_fn(struct v3_vm_info * info, net_state->net_ops = ops; net_state->backend_data = private_data; + net_state->virtio_dev = virtio; + ops->recv = virtio_rx; ops->poll = virtio_nic_poll; - ops->start_tx = virtio_start_tx; - ops->stop_tx = virtio_stop_tx; ops->frontend_data = net_state; + memcpy(ops->fnt_mac, virtio->mac, ETH_ALEN); return 0; } @@ -797,8 +742,7 @@ static int virtio_init(struct v3_vm_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 * dev_id = v3_cfg_val(cfg, "ID"); - - PrintDebug("Virtio NIC: Initializing VIRTIO Network device: %s\n", dev_id); + char * macstr = v3_cfg_val(cfg, "mac"); if (pci_bus == NULL) { PrintError("Virtio NIC: VirtIO devices require a PCI Bus"); @@ -812,14 +756,24 @@ static int virtio_init(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) { virtio_state->pci_bus = pci_bus; virtio_state->vm = vm; - struct vm_device * dev = v3_allocate_device(dev_id, &dev_ops, virtio_state); - if (v3_attach_device(vm, dev) == -1) { + if (macstr != NULL && !str2mac(macstr, virtio_state->mac)) { + PrintDebug("Virtio NIC: Mac specified %s\n", macstr); + }else { + PrintDebug("Virtio NIC: MAC not specified\n"); + random_ethaddr(virtio_state->mac); + } + + struct vm_device * dev = v3_add_device(vm, dev_id, &dev_ops, virtio_state); + + if (dev == NULL) { PrintError("Virtio NIC: Could not attach device %s\n", dev_id); + V3_Free(virtio_state); return -1; } if (v3_dev_add_net_frontend(vm, dev_id, connect_fn, (void *)virtio_state) == -1) { PrintError("Virtio NIC: Could not register %s as net frontend\n", dev_id); + v3_remove_device(dev); return -1; }