* http://www.v3vee.org
*
* Copyright (c) 2008, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2008, Lei Xia <lxia@cs.northwestern.edu>
* Copyright (c) 2008, The V3VEE Project <http://www.v3vee.org>
* All rights reserved.
*
* Author: Jack Lange <jarusl@cs.northwestern.edu>
+ * Lei Xia <lxia@cs.northwestern.edu>
*
* This is free software. You are permitted to use,
* redistribute, and modify it as specified in the file "V3VEE_LICENSE".
#include <palacios/vmm_dev_mgr.h>
#include <palacios/vm_guest_mem.h>
#include <devices/lnx_virtio_pci.h>
-#include <palacios/vmm_vnet.h>
+#include <vnet/vnet.h>
+#include <palacios/vmm_sprintf.h>
#include <devices/pci.h>
-#define QUEUE_SIZE 128
+#ifndef V3_CONFIG_DEBUG_LINUX_VIRTIO_VNET
+#undef PrintDebug
+#define PrintDebug(fmt, args...)
+#endif
+
+
+#define QUEUE_SIZE 4096
+#define CMD_QUEUE_SIZE 128
#define NUM_QUEUES 3
struct vnet_config {
} __attribute__((packed));
+#define CTRL_QUEUE 0
+#define XMIT_QUEUE 1
+#define RECV_QUEUE 2
struct virtio_vnet_state {
+ struct v3_vm_info * vm;
struct vnet_config vnet_cfg;
struct virtio_config virtio_cfg;
-
struct vm_device * pci_bus;
struct pci_device * pci_dev;
-
-
-#define CTRL_QUEUE 0
-#define RECV_QUEUE 1
-#define XMIT_QUEUE 2
+
struct virtio_queue queue[NUM_QUEUES];
struct virtio_queue * cur_queue;
-
int io_range_size;
+ v3_lock_t lock;
+
+ ulong_t pkt_sent, pkt_recv, pkt_drop, tx_exit, rx_exit, total_exit;
+ int ready;
};
#define VNET_GET_ROUTES 10
#define VNET_ADD_LINK 21
#define VNET_DEL_LINK 22
-// structure of the vnet command header
struct vnet_ctrl_hdr {
uint8_t cmd_type;
uint32_t num_cmds;
} __attribute__((packed));
-static int virtio_reset(struct virtio_vnet_state * virtio) {
+struct vnet_bridge_pkt {
+ uint32_t link_id;
+ uint32_t pkt_size;
+ uint8_t pkt[ETHERNET_PACKET_LEN];
+}__attribute__((packed));
+
+
+static int virtio_reset(struct virtio_vnet_state * vnet_state) {
- memset(virtio->queue, 0, sizeof(struct virtio_queue) * 2);
+ memset(vnet_state->queue, 0, sizeof(struct virtio_queue) * NUM_QUEUES);
- virtio->cur_queue = &(virtio->queue[0]);
+ vnet_state->cur_queue = &(vnet_state->queue[0]);
- virtio->virtio_cfg.status = 0;
- virtio->virtio_cfg.pci_isr = 0;
+ vnet_state->virtio_cfg.status = 0;
+ vnet_state->virtio_cfg.pci_isr = 0;
- virtio->queue[0].queue_size = QUEUE_SIZE;
- virtio->queue[1].queue_size = QUEUE_SIZE;
- virtio->queue[2].queue_size = QUEUE_SIZE;
+ vnet_state->queue[0].queue_size = CMD_QUEUE_SIZE;
+ vnet_state->queue[1].queue_size = QUEUE_SIZE;
+ vnet_state->queue[2].queue_size = QUEUE_SIZE;
- memset(&(virtio->vnet_cfg), 0, sizeof(struct vnet_config));
+ memset(&(vnet_state->vnet_cfg), 0, sizeof(struct vnet_config));
+ v3_lock_init(&(vnet_state->lock));
return 0;
}
}
-
-
-static int handle_cmd_kick(struct guest_info * core, struct virtio_vnet_state * vnet_state) {
+static int handle_cmd_kick(struct guest_info * core,
+ struct virtio_vnet_state * vnet_state) {
struct virtio_queue * q = &(vnet_state->queue[0]);
- PrintDebug("VirtioVNET: Virtio Kick on command queue\n");
+ PrintDebug("VNET Bridge: Handling command queue\n");
- while (q->cur_avail_idx < q->avail->index) {
+ while (q->cur_avail_idx != q->avail->index) {
struct vring_desc * hdr_desc = NULL;
struct vring_desc * buf_desc = NULL;
struct vring_desc * status_desc = NULL;
uint8_t * status_ptr = NULL;
uint8_t status = 0;
-
- PrintDebug("Descriptor Count=%d, index=%d, desc_idx=%d\n", desc_cnt, q->cur_avail_idx % QUEUE_SIZE, desc_idx);
+ PrintDebug("VNET Bridge: CMD: Descriptor Count=%d, index=%d, desc_idx=%d\n", desc_cnt, q->cur_avail_idx % QUEUE_SIZE, desc_idx);
if (desc_cnt < 3) {
- PrintError("VirtioVNET cmd must include at least 3 descriptors (cnt=%d)\n", desc_cnt);
+ PrintError("VNET Bridge cmd must include at least 3 descriptors (cnt=%d)\n", desc_cnt);
return -1;
}
hdr_desc = &(q->desc[desc_idx]);
- if (guest_pa_to_host_va(core, hdr_desc->addr_gpa, (addr_t *)&hdr) == -1) {
+ if (v3_gpa_to_hva(core, hdr_desc->addr_gpa, (addr_t *)&hdr) == -1) {
PrintError("Could not translate VirtioVNET header address\n");
return -1;
}
desc_idx = hdr_desc->next;
- if (hdr->cmd_type == VNET_ADD_ROUTE) {
-
+ if (hdr->cmd_type == VNET_ADD_ROUTE) {
for (i = 0; i < hdr->num_cmds; i++) {
uint8_t tmp_status = 0;
struct v3_vnet_route * route = NULL;
buf_desc = &(q->desc[desc_idx]);
- if (guest_pa_to_host_va(core, buf_desc->addr_gpa, (addr_t *)&(route)) == -1) {
+ if (v3_gpa_to_hva(core, buf_desc->addr_gpa, (addr_t *)&(route)) == -1) {
PrintError("Could not translate route address\n");
return -1;
}
- // add route
- PrintDebug("Adding VNET Route\n");
-
tmp_status = v3_vnet_add_route(*route);
-
- PrintDebug("VNET Route Added\n");
-
if (tmp_status != 0) {
PrintError("Error adding VNET ROUTE\n");
+
status = tmp_status;
}
+ PrintDebug("VNET Route Added\n");
+
xfer_len += buf_desc->length;
desc_idx = buf_desc->next;
}
}
-
-
status_desc = &(q->desc[desc_idx]);
- if (guest_pa_to_host_va(core, status_desc->addr_gpa, (addr_t *)&status_ptr) == -1) {
+ if (v3_gpa_to_hva(core, status_desc->addr_gpa, (addr_t *)&status_ptr) == -1) {
PrintError("VirtioVNET Error could not translate status address\n");
return -1;
}
}
-static int virtio_io_write(struct guest_info * core, uint16_t port, void * src, uint_t length, void * private_data) {
+static int vnet_pkt_input_cb(struct v3_vm_info * vm,
+ struct v3_vnet_pkt * pkt,
+ void * private_data){
struct virtio_vnet_state * vnet_state = (struct virtio_vnet_state *)private_data;
- int port_idx = port % vnet_state->io_range_size;
+ struct virtio_queue * q = &(vnet_state->queue[RECV_QUEUE]);
+ int ret_val = -1;
+ unsigned long flags;
+ flags = v3_lock_irqsave(vnet_state->lock);
+
+ if (q->ring_avail_addr == 0) {
+ PrintError("Queue is not set\n");
+
+ goto exit;
+ }
- PrintDebug("VirtioVNET: VIRTIO VNET Write for port %d len=%d, value=%x\n",
- port, length, *(uint32_t *)src);
- PrintDebug("VirtioVNET: port idx=%d\n", port_idx);
+ if (q->cur_avail_idx != q->avail->index) {
+ uint16_t pkt_idx = q->avail->ring[q->cur_avail_idx % q->queue_size];
+ struct vring_desc * pkt_desc = NULL;
+ struct vnet_bridge_pkt * virtio_pkt = NULL;
+
+ pkt_desc = &(q->desc[pkt_idx]);
+
+ if (v3_gpa_to_hva(&(vm->cores[0]), pkt_desc->addr_gpa, (addr_t *)&(virtio_pkt)) == -1) {
+ PrintError("Could not translate buffer address\n");
+ goto exit;
+ }
+
+ PrintDebug("VNET Bridge: RX: pkt sent to guest pkt size: %d, dst link: %d\n", pkt->size, pkt->dst_id);
+
+ // Fill in dst packet buffer
+ virtio_pkt->link_id = pkt->dst_id;
+ virtio_pkt->pkt_size = pkt->size;
+ memcpy(virtio_pkt->pkt, pkt->data, pkt->size);
+
+ 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 = sizeof(struct vnet_bridge_pkt);
+
+ q->used->index++;
+ q->cur_avail_idx++;
+ } else {
+ vnet_state->pkt_drop ++;
+ }
+
+ if (!(q->avail->flags & VIRTIO_NO_IRQ_FLAG)) {
+ v3_pci_raise_irq(vnet_state->pci_bus, 0, vnet_state->pci_dev);
+ vnet_state->virtio_cfg.pci_isr = 0x1;
+ PrintDebug("Raising IRQ %d\n", vnet_state->pci_dev->config_header.intr_line);
+ }
+
+ ret_val = 0;
+
+exit:
+
+ v3_unlock_irqrestore(vnet_state->lock, flags);
+
+ return ret_val;
+}
+
+static int do_tx_pkts(struct guest_info * core,
+ struct virtio_vnet_state * vnet_state)
+{
+ struct virtio_queue * q = &(vnet_state->queue[XMIT_QUEUE]);
+ int recvd = 0;
+
+ if (q->ring_avail_addr == 0) {
+ return -1;
+ }
+
+ while (q->cur_avail_idx != q->avail->index) {
+ uint16_t desc_idx = q->avail->ring[q->cur_avail_idx % q->queue_size];
+ struct vring_desc * pkt_desc = NULL;
+ struct vnet_bridge_pkt * virtio_pkt = NULL;
+
+ pkt_desc = &(q->desc[desc_idx]);
+
+ if (v3_gpa_to_hva(core, pkt_desc->addr_gpa, (addr_t *)&(virtio_pkt)) == -1) {
+ PrintError("Could not translate buffer address\n");
+ return -1;
+ }
+
+ //PrintDebug("VNET Bridge: TX: on cpu %d pkt size: %d, dst link: %d\n", cpu, virtio_pkt->pkt_size, virtio_pkt->link_id);
+
+ struct v3_vnet_pkt pkt;
+ pkt.size = virtio_pkt->pkt_size;
+ pkt.src_type = LINK_EDGE;
+ pkt.src_id = 0;
+ memcpy(pkt.header, virtio_pkt->pkt, ETHERNET_HEADER_LEN);
+ pkt.data = virtio_pkt->pkt;
+ v3_vnet_send_pkt(&pkt, NULL);
+
+ 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 = pkt_desc->length; // What do we set this to????
+ q->used->index++;
+
+ vnet_state->pkt_sent ++;
+ recvd ++;
+
+ q->cur_avail_idx++;
+ }
+
+ if(recvd == 0){
+ return 0;
+ }
+ if (!(q->avail->flags & VIRTIO_NO_IRQ_FLAG)) {
+ v3_pci_raise_irq(vnet_state->pci_bus, 0, vnet_state->pci_dev);
+ vnet_state->virtio_cfg.pci_isr = 0x1;
+ }
+
+ return 0;
+}
+
+static void vnet_virtio_poll(struct v3_vm_info * vm, void * private_data){
+ struct virtio_vnet_state * vnet_state = (struct virtio_vnet_state *)private_data;
+
+ if(vm == vnet_state->vm){
+ do_tx_pkts(&(vm->cores[0]), vnet_state);
+ }
+}
+
+static int handle_rx_queue_kick(struct guest_info *core,
+ struct virtio_vnet_state * vnet_state)
+{
+ return 0;
+}
+
+static int vnet_virtio_io_write(struct guest_info * core,
+ uint16_t port, void * src,
+ uint_t length, void * private_data) {
+ struct virtio_vnet_state * vnet_state = (struct virtio_vnet_state *)private_data;
+ int port_idx = port % vnet_state->io_range_size;
+
+ PrintDebug("VNET Bridge: VIRTIO VNET Write for port %d len=%d, value=%x\n",
+ port, length, *(uint32_t *)src);
+
+ vnet_state->total_exit ++;
switch (port_idx) {
case GUEST_FEATURES_PORT:
if (length != 4) {
PrintError("Illegal write length for guest features\n");
return -1;
- }
-
+ }
vnet_state->virtio_cfg.guest_features = *(uint32_t *)src;
break;
// round up to next page boundary.
vnet_state->cur_queue->ring_used_addr = (vnet_state->cur_queue->ring_used_addr + 0xfff) & ~0xfff;
- if (guest_pa_to_host_va(core, vnet_state->cur_queue->ring_desc_addr, (addr_t *)&(vnet_state->cur_queue->desc)) == -1) {
+ if (v3_gpa_to_hva(core, vnet_state->cur_queue->ring_desc_addr, (addr_t *)&(vnet_state->cur_queue->desc)) == -1) {
PrintError("Could not translate ring descriptor address\n");
return -1;
}
-
- if (guest_pa_to_host_va(core, vnet_state->cur_queue->ring_avail_addr, (addr_t *)&(vnet_state->cur_queue->avail)) == -1) {
+ if (v3_gpa_to_hva(core, vnet_state->cur_queue->ring_avail_addr, (addr_t *)&(vnet_state->cur_queue->avail)) == -1) {
PrintError("Could not translate ring available address\n");
return -1;
}
-
- if (guest_pa_to_host_va(core, vnet_state->cur_queue->ring_used_addr, (addr_t *)&(vnet_state->cur_queue->used)) == -1) {
+ if (v3_gpa_to_hva(core, vnet_state->cur_queue->ring_used_addr, (addr_t *)&(vnet_state->cur_queue->used)) == -1) {
PrintError("Could not translate ring used address\n");
return -1;
}
- PrintDebug("VirtioVNET: RingDesc_addr=%p, Avail_addr=%p, Used_addr=%p\n",
+ PrintDebug("VNET Bridge: RingDesc_addr=%p, Avail_addr=%p, Used_addr=%p\n",
(void *)(vnet_state->cur_queue->ring_desc_addr),
(void *)(vnet_state->cur_queue->ring_avail_addr),
(void *)(vnet_state->cur_queue->ring_used_addr));
- PrintDebug("VirtioVNET: RingDesc=%p, Avail=%p, Used=%p\n",
+ PrintDebug("VNET Bridge: RingDesc=%p, Avail=%p, Used=%p\n",
vnet_state->cur_queue->desc, vnet_state->cur_queue->avail, vnet_state->cur_queue->used);
+ if(vnet_state->queue[RECV_QUEUE].avail != NULL){
+ vnet_state->ready = 1;
+ vnet_state->queue[RECV_QUEUE].used->flags |= VRING_NO_NOTIFY_FLAG;
+ }
+
+ //No notify when there is pkt tx from guest
+ //palacios will do the polling
+ if(vnet_state->queue[XMIT_QUEUE].used != NULL){
+ vnet_state->queue[XMIT_QUEUE].used->flags |= VRING_NO_NOTIFY_FLAG;
+ }
} else {
PrintError("Illegal write length for page frame number\n");
return -1;
vnet_state->virtio_cfg.vring_queue_selector = *(uint16_t *)src;
if (vnet_state->virtio_cfg.vring_queue_selector > NUM_QUEUES) {
- PrintError("Virtio Symbiotic device has no qeueues. Selected %d\n",
+ PrintError("VNET Bridge device has no qeueues. Selected %d\n",
vnet_state->virtio_cfg.vring_queue_selector);
return -1;
}
case VRING_Q_NOTIFY_PORT: {
uint16_t queue_idx = *(uint16_t *)src;
- PrintDebug("VirtioVNET: Handling Kick\n");
+ PrintDebug("VNET Bridge: Handling Kick\n");
if (queue_idx == 0) {
if (handle_cmd_kick(core, vnet_state) == -1) {
- PrintError("Could not handle VNET Control command\n");
+ PrintError("Could not handle Virtio VNET Control command\n");
return -1;
}
} else if (queue_idx == 1) {
-
- // down queue
+ if (do_tx_pkts(core, vnet_state) == -1){
+ PrintError("Could not handle Virtio VNET TX\n");
+ return -1;
+ }
+ PrintError("Notify on TX\n");
} else if (queue_idx == 2) {
- // up queue
+ if (handle_rx_queue_kick(core, vnet_state) == -1){
+ PrintError("Could not handle Virtio RX buffer refills Kick\n");
+ return -1;
+ }
+ vnet_state->rx_exit ++;
} else {
- PrintError("Kick on invalid queue (%d)\n", queue_idx);
+ PrintError("VNET Bridge: Kick on invalid queue (%d)\n", queue_idx);
return -1;
}
vnet_state->virtio_cfg.status = *(uint8_t *)src;
if (vnet_state->virtio_cfg.status == 0) {
- PrintDebug("VirtioVNET: Resetting device\n");
+ PrintDebug("VNET Bridge: Resetting device\n");
virtio_reset(vnet_state);
}
}
-static int virtio_io_read(struct guest_info * core, uint16_t port, void * dst, uint_t length, void * private_data) {
+static int vnet_virtio_io_read(struct guest_info * core,
+ uint16_t port, void * dst,
+ uint_t length, void * private_data) {
struct virtio_vnet_state * vnet_state = (struct virtio_vnet_state *)private_data;
int port_idx = port % vnet_state->io_range_size;
-/*
- PrintDebug("VirtioVNET: VIRTIO SYMBIOTIC Read for port %d (index =%d), length=%d\n",
- port, port_idx, length);
-*/
switch (port_idx) {
case HOST_FEATURES_PORT:
if (length != 4) {
}
+static int virtio_free(struct virtio_vnet_state * vnet_state) {
+
+ // unregister from PCI
+
+ V3_Free(vnet_state);
+ return 0;
+}
+
static struct v3_device_ops dev_ops = {
- .free = NULL,
- .reset = NULL,
- .start = NULL,
- .stop = NULL,
+ .free = (int (*)(void *))virtio_free,
};
-
-static int vnet_init(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
+static int dev_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_vnet_state * virtio_state = NULL;
+ struct virtio_vnet_state * vnet_state = NULL;
struct pci_device * pci_dev = NULL;
char * name = v3_cfg_val(cfg, "name");
- PrintDebug("VirtioVNET: Initializing VIRTIO VNET Control device\n");
+ PrintDebug("VNET Bridge: Initializing VNET Bridge Control device: %s\n", name);
if (pci_bus == NULL) {
- PrintError("VirtIO devices require a PCI Bus");
+ PrintError("VNET Bridge device require a PCI Bus");
return -1;
}
- virtio_state = (struct virtio_vnet_state *)V3_Malloc(sizeof(struct virtio_vnet_state));
- memset(virtio_state, 0, sizeof(struct virtio_vnet_state));
+ vnet_state = (struct virtio_vnet_state *)V3_Malloc(sizeof(struct virtio_vnet_state));
+ memset(vnet_state, 0, sizeof(struct virtio_vnet_state));
+
+ vnet_state->vm = vm;
- struct vm_device * dev = v3_allocate_device(name, &dev_ops, virtio_state);
+ struct vm_device * dev = v3_add_device(vm, name, &dev_ops, vnet_state);
- if (v3_attach_device(vm, dev) == -1) {
+ if (dev == NULL) {
PrintError("Could not attach device %s\n", name);
+ V3_Free(vnet_state);
return -1;
}
int tmp_ports = num_ports;
int i;
-
// 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
+ vnet_state->io_range_size = 1; // must be a power of 2
while (tmp_ports > 0) {
tmp_ports >>= 1;
- virtio_state->io_range_size <<= 1;
+ vnet_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;
+ if ((num_ports & ((vnet_state->io_range_size >> 1) - 1)) == 0) {
+ vnet_state->io_range_size >>= 1;
}
-
for (i = 0; i < 6; i++) {
bars[i].type = PCI_BAR_NONE;
}
bars[0].type = PCI_BAR_IO;
bars[0].default_base_port = -1;
- bars[0].num_ports = virtio_state->io_range_size;
-
- bars[0].io_read = virtio_io_read;
- bars[0].io_write = virtio_io_write;
- bars[0].private_data = virtio_state;
+ bars[0].num_ports = vnet_state->io_range_size;
+ bars[0].io_read = vnet_virtio_io_read;
+ bars[0].io_write = vnet_virtio_io_write;
+ bars[0].private_data = vnet_state;
pci_dev = v3_pci_register_device(pci_bus, PCI_STD_DEVICE,
- 0, PCI_AUTO_DEV_NUM, 0,
+ 0, 5 /*PCI_AUTO_DEV_NUM*/, 0,
"LNX_VIRTIO_VNET", bars,
- NULL, NULL, NULL, virtio_state);
+ NULL, NULL, NULL, vnet_state);
if (!pci_dev) {
PrintError("Could not register PCI Device\n");
+ v3_remove_device(dev);
return -1;
}
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_VNET_DEV_ID;
pci_dev->config_header.class = PCI_CLASS_MEMORY;
pci_dev->config_header.subclass = PCI_MEM_SUBCLASS_RAM;
-
pci_dev->config_header.subsystem_id = VIRTIO_VNET_SUBDEVICE_ID;
-
-
pci_dev->config_header.intr_pin = 1;
-
pci_dev->config_header.max_latency = 1; // ?? (qemu does it...)
- virtio_state->pci_dev = pci_dev;
- virtio_state->pci_bus = pci_bus;
+ vnet_state->pci_dev = pci_dev;
+ vnet_state->pci_bus = pci_bus;
}
- virtio_reset(virtio_state);
+ virtio_reset(vnet_state);
+
+ struct v3_vnet_bridge_ops brg_ops;
+ brg_ops.input = vnet_pkt_input_cb;
+ brg_ops.poll = vnet_virtio_poll;
+
+ V3_Print("Registering Virtio device as vnet bridge\n");
+
+ v3_vnet_add_bridge(vm, &brg_ops, CTL_VM_BRIDGE, (void *)vnet_state);
-
return 0;
}
-device_register("LNX_VIRTIO_VNET", vnet_init)
+device_register("LNX_VIRTIO_VNET", dev_init)
