From c8edafdabc27533d9c51eddc2896e772c16d965c Mon Sep 17 00:00:00 2001 From: vmaffione Date: Sun, 7 Jul 2019 12:15:24 +0000 Subject: [PATCH] bhyve: abstraction for network backends Bhyve can currently emulate two virtual NICs, namely virtio-net and e1000, and connect to the host network through two backends, namely tap and netmap. However, there is no interface between virtual NIC functionalities and backend functionalities. As a result, the backend code is duplicated between the two virtual NIC implementations and also within the same virtual NIC. Also, e1000 cannot currently use netmap as a backend. This patch introduces a network backend API between virtio-net/e1000 and tap/netmap, to improve code reuse and add missing functionalities. Virtual NICs and backends can negotiate virtio-net features, such as checksum offload and TSO. If the backend supports the features, it will propagate this information to the guest, so that the latter can make use of them. Currently, only netmap VALE ports support the features, but support should be added to tap in the future. Reviewed by: jhb, bryanv MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D20659 --- usr.sbin/bhyve/Makefile | 1 + usr.sbin/bhyve/net_backends.c | 806 ++++++++++++++++++++++++++++++++ usr.sbin/bhyve/net_backends.h | 87 ++++ usr.sbin/bhyve/pci_e82545.c | 83 +--- usr.sbin/bhyve/pci_virtio_net.c | 538 +++------------------ 5 files changed, 979 insertions(+), 536 deletions(-) create mode 100644 usr.sbin/bhyve/net_backends.c create mode 100644 usr.sbin/bhyve/net_backends.h diff --git a/usr.sbin/bhyve/Makefile b/usr.sbin/bhyve/Makefile index e203a57c18a..2c38e55b643 100644 --- a/usr.sbin/bhyve/Makefile +++ b/usr.sbin/bhyve/Makefile @@ -34,6 +34,7 @@ SRCS= \ mem.c \ mevent.c \ mptbl.c \ + net_backends.c \ net_utils.c \ pci_ahci.c \ pci_e82545.c \ diff --git a/usr.sbin/bhyve/net_backends.c b/usr.sbin/bhyve/net_backends.c new file mode 100644 index 00000000000..be94b4f5d54 --- /dev/null +++ b/usr.sbin/bhyve/net_backends.c @@ -0,0 +1,806 @@ +/*- + * Copyright (c) 2019 Vincenzo Maffione + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR + * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS + * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, + * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT + * OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR + * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE + * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, + * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * $FreeBSD$ + */ + +/* + * This file implements multiple network backends (tap, netmap, ...), + * to be used by network frontends such as virtio-net and e1000. + * The API to access the backend (e.g. send/receive packets, negotiate + * features) is exported by net_backends.h. + */ + +#include /* u_short etc */ +#ifndef WITHOUT_CAPSICUM +#include +#endif +#include +#include +#include +#include + +#include +#include +#include +#define NETMAP_WITH_LIBS +#include + +#ifndef WITHOUT_CAPSICUM +#include +#endif +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +#include "iov.h" +#include "mevent.h" +#include "net_backends.h" + +#include + +/* + * Each network backend registers a set of function pointers that are + * used to implement the net backends API. + * This might need to be exposed if we implement backends in separate files. + */ +struct net_backend { + const char *prefix; /* prefix matching this backend */ + + /* + * Routines used to initialize and cleanup the resources needed + * by a backend. The cleanup function is used internally, + * and should not be called by the frontend. + */ + int (*init)(struct net_backend *be, const char *devname, + net_be_rxeof_t cb, void *param); + void (*cleanup)(struct net_backend *be); + + /* + * Called to serve a guest transmit request. The scatter-gather + * vector provided by the caller has 'iovcnt' elements and contains + * the packet to send. + */ + ssize_t (*send)(struct net_backend *be, struct iovec *iov, int iovcnt); + + /* + * Called to receive a packet from the backend. When the function + * returns a positive value 'len', the scatter-gather vector + * provided by the caller contains a packet with such length. + * The function returns 0 if the backend doesn't have a new packet to + * receive. + */ + ssize_t (*recv)(struct net_backend *be, struct iovec *iov, int iovcnt); + + /* + * Ask the backend for the virtio-net features it is able to + * support. Possible features are TSO, UFO and checksum offloading + * in both rx and tx direction and for both IPv4 and IPv6. + */ + uint64_t (*get_cap)(struct net_backend *be); + + /* + * Tell the backend to enable/disable the specified virtio-net + * features (capabilities). + */ + int (*set_cap)(struct net_backend *be, uint64_t features, + unsigned int vnet_hdr_len); + + struct pci_vtnet_softc *sc; + int fd; + + /* + * Length of the virtio-net header used by the backend and the + * frontend, respectively. A zero value means that the header + * is not used. + */ + unsigned int be_vnet_hdr_len; + unsigned int fe_vnet_hdr_len; + + /* Size of backend-specific private data. */ + size_t priv_size; + + /* Room for backend-specific data. */ + char opaque[0]; +}; + +SET_DECLARE(net_backend_set, struct net_backend); + +#define VNET_HDR_LEN sizeof(struct virtio_net_rxhdr) + +#define WPRINTF(params) printf params + +/* + * The tap backend + */ + +struct tap_priv { + struct mevent *mevp; +}; + +static void +tap_cleanup(struct net_backend *be) +{ + struct tap_priv *priv = (struct tap_priv *)be->opaque; + + if (priv->mevp) { + mevent_delete(priv->mevp); + } + if (be->fd != -1) { + close(be->fd); + be->fd = -1; + } +} + +static int +tap_init(struct net_backend *be, const char *devname, + net_be_rxeof_t cb, void *param) +{ + struct tap_priv *priv = (struct tap_priv *)be->opaque; + char tbuf[80]; + int fd; + int opt = 1; +#ifndef WITHOUT_CAPSICUM + cap_rights_t rights; +#endif + + if (cb == NULL) { + WPRINTF(("TAP backend requires non-NULL callback\n")); + return (-1); + } + + strcpy(tbuf, "/dev/"); + strlcat(tbuf, devname, sizeof(tbuf)); + + fd = open(tbuf, O_RDWR); + if (fd == -1) { + WPRINTF(("open of tap device %s failed\n", tbuf)); + goto error; + } + + /* + * Set non-blocking and register for read + * notifications with the event loop + */ + if (ioctl(fd, FIONBIO, &opt) < 0) { + WPRINTF(("tap device O_NONBLOCK failed\n")); + goto error; + } + +#ifndef WITHOUT_CAPSICUM + cap_rights_init(&rights, CAP_EVENT, CAP_READ, CAP_WRITE); + if (caph_rights_limit(fd, &rights) == -1) + errx(EX_OSERR, "Unable to apply rights for sandbox"); +#endif + + priv->mevp = mevent_add(fd, EVF_READ, cb, param); + if (priv->mevp == NULL) { + WPRINTF(("Could not register event\n")); + goto error; + } + + be->fd = fd; + + return (0); + +error: + tap_cleanup(be); + return (-1); +} + +/* + * Called to send a buffer chain out to the tap device + */ +static ssize_t +tap_send(struct net_backend *be, struct iovec *iov, int iovcnt) +{ + return (writev(be->fd, iov, iovcnt)); +} + +static ssize_t +tap_recv(struct net_backend *be, struct iovec *iov, int iovcnt) +{ + ssize_t ret; + + /* Should never be called without a valid tap fd */ + assert(be->fd != -1); + + ret = readv(be->fd, iov, iovcnt); + + if (ret < 0 && errno == EWOULDBLOCK) { + return (0); + } + + return (ret); +} + +static uint64_t +tap_get_cap(struct net_backend *be) +{ + + return (0); /* no capabilities for now */ +} + +static int +tap_set_cap(struct net_backend *be, uint64_t features, + unsigned vnet_hdr_len) +{ + + return ((features || vnet_hdr_len) ? -1 : 0); +} + +static struct net_backend tap_backend = { + .prefix = "tap", + .priv_size = sizeof(struct tap_priv), + .init = tap_init, + .cleanup = tap_cleanup, + .send = tap_send, + .recv = tap_recv, + .get_cap = tap_get_cap, + .set_cap = tap_set_cap, +}; + +/* A clone of the tap backend, with a different prefix. */ +static struct net_backend vmnet_backend = { + .prefix = "vmnet", + .priv_size = sizeof(struct tap_priv), + .init = tap_init, + .cleanup = tap_cleanup, + .send = tap_send, + .recv = tap_recv, + .get_cap = tap_get_cap, + .set_cap = tap_set_cap, +}; + +DATA_SET(net_backend_set, tap_backend); +DATA_SET(net_backend_set, vmnet_backend); + +/* + * The netmap backend + */ + +/* The virtio-net features supported by netmap. */ +#define NETMAP_FEATURES (VIRTIO_NET_F_CSUM | VIRTIO_NET_F_HOST_TSO4 | \ + VIRTIO_NET_F_HOST_TSO6 | VIRTIO_NET_F_HOST_UFO | \ + VIRTIO_NET_F_GUEST_CSUM | VIRTIO_NET_F_GUEST_TSO4 | \ + VIRTIO_NET_F_GUEST_TSO6 | VIRTIO_NET_F_GUEST_UFO) + +struct netmap_priv { + char ifname[IFNAMSIZ]; + struct nm_desc *nmd; + uint16_t memid; + struct netmap_ring *rx; + struct netmap_ring *tx; + struct mevent *mevp; + net_be_rxeof_t cb; + void *cb_param; +}; + +static void +nmreq_init(struct nmreq *req, char *ifname) +{ + + memset(req, 0, sizeof(*req)); + strlcpy(req->nr_name, ifname, sizeof(req->nr_name)); + req->nr_version = NETMAP_API; +} + +static int +netmap_set_vnet_hdr_len(struct net_backend *be, int vnet_hdr_len) +{ + int err; + struct nmreq req; + struct netmap_priv *priv = (struct netmap_priv *)be->opaque; + + nmreq_init(&req, priv->ifname); + req.nr_cmd = NETMAP_BDG_VNET_HDR; + req.nr_arg1 = vnet_hdr_len; + err = ioctl(be->fd, NIOCREGIF, &req); + if (err) { + WPRINTF(("Unable to set vnet header length %d\n", + vnet_hdr_len)); + return (err); + } + + be->be_vnet_hdr_len = vnet_hdr_len; + + return (0); +} + +static int +netmap_has_vnet_hdr_len(struct net_backend *be, unsigned vnet_hdr_len) +{ + int prev_hdr_len = be->be_vnet_hdr_len; + int ret; + + if (vnet_hdr_len == prev_hdr_len) { + return (1); + } + + ret = netmap_set_vnet_hdr_len(be, vnet_hdr_len); + if (ret) { + return (0); + } + + netmap_set_vnet_hdr_len(be, prev_hdr_len); + + return (1); +} + +static uint64_t +netmap_get_cap(struct net_backend *be) +{ + + return (netmap_has_vnet_hdr_len(be, VNET_HDR_LEN) ? + NETMAP_FEATURES : 0); +} + +static int +netmap_set_cap(struct net_backend *be, uint64_t features, + unsigned vnet_hdr_len) +{ + + return (netmap_set_vnet_hdr_len(be, vnet_hdr_len)); +} + +static int +netmap_init(struct net_backend *be, const char *devname, + net_be_rxeof_t cb, void *param) +{ + struct netmap_priv *priv = (struct netmap_priv *)be->opaque; + + strlcpy(priv->ifname, devname, sizeof(priv->ifname)); + priv->ifname[sizeof(priv->ifname) - 1] = '\0'; + + priv->nmd = nm_open(priv->ifname, NULL, NETMAP_NO_TX_POLL, NULL); + if (priv->nmd == NULL) { + WPRINTF(("Unable to nm_open(): interface '%s', errno (%s)\n", + devname, strerror(errno))); + free(priv); + return (-1); + } + + priv->memid = priv->nmd->req.nr_arg2; + priv->tx = NETMAP_TXRING(priv->nmd->nifp, 0); + priv->rx = NETMAP_RXRING(priv->nmd->nifp, 0); + priv->cb = cb; + priv->cb_param = param; + be->fd = priv->nmd->fd; + + priv->mevp = mevent_add(be->fd, EVF_READ, cb, param); + if (priv->mevp == NULL) { + WPRINTF(("Could not register event\n")); + return (-1); + } + + return (0); +} + +static void +netmap_cleanup(struct net_backend *be) +{ + struct netmap_priv *priv = (struct netmap_priv *)be->opaque; + + if (priv->mevp) { + mevent_delete(priv->mevp); + } + if (priv->nmd) { + nm_close(priv->nmd); + } + be->fd = -1; +} + +static ssize_t +netmap_send(struct net_backend *be, struct iovec *iov, + int iovcnt) +{ + struct netmap_priv *priv = (struct netmap_priv *)be->opaque; + struct netmap_ring *ring; + ssize_t totlen = 0; + int nm_buf_size; + int nm_buf_len; + uint32_t head; + void *nm_buf; + int j; + + ring = priv->tx; + head = ring->head; + if (head == ring->tail) { + WPRINTF(("No space, drop %zu bytes\n", count_iov(iov, iovcnt))); + goto txsync; + } + nm_buf = NETMAP_BUF(ring, ring->slot[head].buf_idx); + nm_buf_size = ring->nr_buf_size; + nm_buf_len = 0; + + for (j = 0; j < iovcnt; j++) { + int iov_frag_size = iov[j].iov_len; + void *iov_frag_buf = iov[j].iov_base; + + totlen += iov_frag_size; + + /* + * Split each iovec fragment over more netmap slots, if + * necessary. + */ + for (;;) { + int copylen; + + copylen = iov_frag_size < nm_buf_size ? iov_frag_size : nm_buf_size; + memcpy(nm_buf, iov_frag_buf, copylen); + + iov_frag_buf += copylen; + iov_frag_size -= copylen; + nm_buf += copylen; + nm_buf_size -= copylen; + nm_buf_len += copylen; + + if (iov_frag_size == 0) { + break; + } + + ring->slot[head].len = nm_buf_len; + ring->slot[head].flags = NS_MOREFRAG; + head = nm_ring_next(ring, head); + if (head == ring->tail) { + /* + * We ran out of netmap slots while + * splitting the iovec fragments. + */ + WPRINTF(("No space, drop %zu bytes\n", + count_iov(iov, iovcnt))); + goto txsync; + } + nm_buf = NETMAP_BUF(ring, ring->slot[head].buf_idx); + nm_buf_size = ring->nr_buf_size; + nm_buf_len = 0; + } + } + + /* Complete the last slot, which must not have NS_MOREFRAG set. */ + ring->slot[head].len = nm_buf_len; + ring->slot[head].flags = 0; + head = nm_ring_next(ring, head); + + /* Now update ring->head and ring->cur. */ + ring->head = ring->cur = head; +txsync: + ioctl(be->fd, NIOCTXSYNC, NULL); + + return (totlen); +} + +static ssize_t +netmap_recv(struct net_backend *be, struct iovec *iov, int iovcnt) +{ + struct netmap_priv *priv = (struct netmap_priv *)be->opaque; + struct netmap_slot *slot = NULL; + struct netmap_ring *ring; + void *iov_frag_buf; + int iov_frag_size; + ssize_t totlen = 0; + uint32_t head; + + assert(iovcnt); + + ring = priv->rx; + head = ring->head; + iov_frag_buf = iov->iov_base; + iov_frag_size = iov->iov_len; + + do { + int nm_buf_len; + void *nm_buf; + + if (head == ring->tail) { + return (0); + } + + slot = ring->slot + head; + nm_buf = NETMAP_BUF(ring, slot->buf_idx); + nm_buf_len = slot->len; + + for (;;) { + int copylen = nm_buf_len < iov_frag_size ? + nm_buf_len : iov_frag_size; + + memcpy(iov_frag_buf, nm_buf, copylen); + nm_buf += copylen; + nm_buf_len -= copylen; + iov_frag_buf += copylen; + iov_frag_size -= copylen; + totlen += copylen; + + if (nm_buf_len == 0) { + break; + } + + iov++; + iovcnt--; + if (iovcnt == 0) { + /* No space to receive. */ + WPRINTF(("Short iov, drop %zd bytes\n", + totlen)); + return (-ENOSPC); + } + iov_frag_buf = iov->iov_base; + iov_frag_size = iov->iov_len; + } + + head = nm_ring_next(ring, head); + + } while (slot->flags & NS_MOREFRAG); + + /* Release slots to netmap. */ + ring->head = ring->cur = head; + + return (totlen); +} + +static struct net_backend netmap_backend = { + .prefix = "netmap", + .priv_size = sizeof(struct netmap_priv), + .init = netmap_init, + .cleanup = netmap_cleanup, + .send = netmap_send, + .recv = netmap_recv, + .get_cap = netmap_get_cap, + .set_cap = netmap_set_cap, +}; + +/* A clone of the netmap backend, with a different prefix. */ +static struct net_backend vale_backend = { + .prefix = "vale", + .priv_size = sizeof(struct netmap_priv), + .init = netmap_init, + .cleanup = netmap_cleanup, + .send = netmap_send, + .recv = netmap_recv, + .get_cap = netmap_get_cap, + .set_cap = netmap_set_cap, +}; + +DATA_SET(net_backend_set, netmap_backend); +DATA_SET(net_backend_set, vale_backend); + +/* + * Initialize a backend and attach to the frontend. + * This is called during frontend initialization. + * @pbe is a pointer to the backend to be initialized + * @devname is the backend-name as supplied on the command line, + * e.g. -s 2:0,frontend-name,backend-name[,other-args] + * @cb is the receive callback supplied by the frontend, + * and it is invoked in the event loop when a receive + * event is generated in the hypervisor, + * @param is a pointer to the frontend, and normally used as + * the argument for the callback. + */ +int +netbe_init(struct net_backend **ret, const char *devname, net_be_rxeof_t cb, + void *param) +{ + struct net_backend **pbe, *nbe, *tbe = NULL; + int err; + + /* + * Find the network backend that matches the user-provided + * device name. net_backend_set is built using a linker set. + */ + SET_FOREACH(pbe, net_backend_set) { + if (strncmp(devname, (*pbe)->prefix, + strlen((*pbe)->prefix)) == 0) { + tbe = *pbe; + assert(tbe->init != NULL); + assert(tbe->cleanup != NULL); + assert(tbe->send != NULL); + assert(tbe->recv != NULL); + assert(tbe->get_cap != NULL); + assert(tbe->set_cap != NULL); + break; + } + } + + *ret = NULL; + if (tbe == NULL) + return (EINVAL); + nbe = calloc(1, sizeof(*nbe) + tbe->priv_size); + *nbe = *tbe; /* copy the template */ + nbe->fd = -1; + nbe->sc = param; + nbe->be_vnet_hdr_len = 0; + nbe->fe_vnet_hdr_len = 0; + + /* Initialize the backend. */ + err = nbe->init(nbe, devname, cb, param); + if (err) { + free(nbe); + return (err); + } + + *ret = nbe; + + return (0); +} + +void +netbe_cleanup(struct net_backend *be) +{ + + if (be != NULL) { + be->cleanup(be); + free(be); + } +} + +uint64_t +netbe_get_cap(struct net_backend *be) +{ + + assert(be != NULL); + return (be->get_cap(be)); +} + +int +netbe_set_cap(struct net_backend *be, uint64_t features, + unsigned vnet_hdr_len) +{ + int ret; + + assert(be != NULL); + + /* There are only three valid lengths, i.e., 0, 10 and 12. */ + if (vnet_hdr_len && vnet_hdr_len != VNET_HDR_LEN + && vnet_hdr_len != (VNET_HDR_LEN - sizeof(uint16_t))) + return (-1); + + be->fe_vnet_hdr_len = vnet_hdr_len; + + ret = be->set_cap(be, features, vnet_hdr_len); + assert(be->be_vnet_hdr_len == 0 || + be->be_vnet_hdr_len == be->fe_vnet_hdr_len); + + return (ret); +} + +static __inline struct iovec * +iov_trim(struct iovec *iov, int *iovcnt, unsigned int tlen) +{ + struct iovec *riov; + + /* XXX short-cut: assume first segment is >= tlen */ + assert(iov[0].iov_len >= tlen); + + iov[0].iov_len -= tlen; + if (iov[0].iov_len == 0) { + assert(*iovcnt > 1); + *iovcnt -= 1; + riov = &iov[1]; + } else { + iov[0].iov_base = (void *)((uintptr_t)iov[0].iov_base + tlen); + riov = &iov[0]; + } + + return (riov); +} + +ssize_t +netbe_send(struct net_backend *be, struct iovec *iov, int iovcnt) +{ + + assert(be != NULL); + if (be->be_vnet_hdr_len != be->fe_vnet_hdr_len) { + /* + * The frontend uses a virtio-net header, but the backend + * does not. We ignore it (as it must be all zeroes) and + * strip it. + */ + assert(be->be_vnet_hdr_len == 0); + iov = iov_trim(iov, &iovcnt, be->fe_vnet_hdr_len); + } + + return (be->send(be, iov, iovcnt)); +} + +/* + * Try to read a packet from the backend, without blocking. + * If no packets are available, return 0. In case of success, return + * the length of the packet just read. Return -1 in case of errors. + */ +ssize_t +netbe_recv(struct net_backend *be, struct iovec *iov, int iovcnt) +{ + /* Length of prepended virtio-net header. */ + unsigned int hlen = be->fe_vnet_hdr_len; + int ret; + + assert(be != NULL); + + if (hlen && hlen != be->be_vnet_hdr_len) { + /* + * The frontend uses a virtio-net header, but the backend + * does not. We need to prepend a zeroed header. + */ + struct virtio_net_rxhdr *vh; + + assert(be->be_vnet_hdr_len == 0); + + /* + * Get a pointer to the rx header, and use the + * data immediately following it for the packet buffer. + */ + vh = iov[0].iov_base; + iov = iov_trim(iov, &iovcnt, hlen); + + /* + * The only valid field in the rx packet header is the + * number of buffers if merged rx bufs were negotiated. + */ + memset(vh, 0, hlen); + if (hlen == VNET_HDR_LEN) { + vh->vrh_bufs = 1; + } + } + + ret = be->recv(be, iov, iovcnt); + if (ret > 0) { + ret += hlen; + } + + return (ret); +} + +/* + * Read a packet from the backend and discard it. + * Returns the size of the discarded packet or zero if no packet was available. + * A negative error code is returned in case of read error. + */ +ssize_t +netbe_rx_discard(struct net_backend *be) +{ + /* + * MP note: the dummybuf is only used to discard frames, + * so there is no need for it to be per-vtnet or locked. + * We only make it large enough for TSO-sized segment. + */ + static uint8_t dummybuf[65536 + 64]; + struct iovec iov; + + iov.iov_base = dummybuf; + iov.iov_len = sizeof(dummybuf); + + return netbe_recv(be, &iov, 1); +} + diff --git a/usr.sbin/bhyve/net_backends.h b/usr.sbin/bhyve/net_backends.h new file mode 100644 index 00000000000..0b6311e38a6 --- /dev/null +++ b/usr.sbin/bhyve/net_backends.h @@ -0,0 +1,87 @@ +/*- + * Copyright (c) 2019 Vincenzo Maffione + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR + * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS + * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, + * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT + * OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR + * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE + * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, + * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * $FreeBSD$ + */ + +#ifndef __NET_BACKENDS_H__ +#define __NET_BACKENDS_H__ + +#include + +/* Opaque type representing a network backend. */ +typedef struct net_backend net_backend_t; + +/* Interface between network frontends and the network backends. */ +typedef void (*net_be_rxeof_t)(int, enum ev_type, void *param); +int netbe_init(net_backend_t **be, const char *devname, net_be_rxeof_t cb, + void *param); +void netbe_cleanup(net_backend_t *be); +uint64_t netbe_get_cap(net_backend_t *be); +int netbe_set_cap(net_backend_t *be, uint64_t cap, + unsigned vnet_hdr_len); +ssize_t netbe_send(net_backend_t *be, struct iovec *iov, int iovcnt); +ssize_t netbe_recv(net_backend_t *be, struct iovec *iov, int iovcnt); +ssize_t netbe_rx_discard(net_backend_t *be); + + +/* + * Network device capabilities taken from the VirtIO standard. + * Despite the name, these capabilities can be used by different frontents + * (virtio-net, ptnet) and supported by different backends (netmap, tap, ...). + */ +#define VIRTIO_NET_F_CSUM (1 << 0) /* host handles partial cksum */ +#define VIRTIO_NET_F_GUEST_CSUM (1 << 1) /* guest handles partial cksum */ +#define VIRTIO_NET_F_MAC (1 << 5) /* host supplies MAC */ +#define VIRTIO_NET_F_GSO_DEPREC (1 << 6) /* deprecated: host handles GSO */ +#define VIRTIO_NET_F_GUEST_TSO4 (1 << 7) /* guest can rcv TSOv4 */ +#define VIRTIO_NET_F_GUEST_TSO6 (1 << 8) /* guest can rcv TSOv6 */ +#define VIRTIO_NET_F_GUEST_ECN (1 << 9) /* guest can rcv TSO with ECN */ +#define VIRTIO_NET_F_GUEST_UFO (1 << 10) /* guest can rcv UFO */ +#define VIRTIO_NET_F_HOST_TSO4 (1 << 11) /* host can rcv TSOv4 */ +#define VIRTIO_NET_F_HOST_TSO6 (1 << 12) /* host can rcv TSOv6 */ +#define VIRTIO_NET_F_HOST_ECN (1 << 13) /* host can rcv TSO with ECN */ +#define VIRTIO_NET_F_HOST_UFO (1 << 14) /* host can rcv UFO */ +#define VIRTIO_NET_F_MRG_RXBUF (1 << 15) /* host can merge RX buffers */ +#define VIRTIO_NET_F_STATUS (1 << 16) /* config status field available */ +#define VIRTIO_NET_F_CTRL_VQ (1 << 17) /* control channel available */ +#define VIRTIO_NET_F_CTRL_RX (1 << 18) /* control channel RX mode support */ +#define VIRTIO_NET_F_CTRL_VLAN (1 << 19) /* control channel VLAN filtering */ +#define VIRTIO_NET_F_GUEST_ANNOUNCE \ + (1 << 21) /* guest can send gratuitous pkts */ + +/* + * Fixed network header size + */ +struct virtio_net_rxhdr { + uint8_t vrh_flags; + uint8_t vrh_gso_type; + uint16_t vrh_hdr_len; + uint16_t vrh_gso_size; + uint16_t vrh_csum_start; + uint16_t vrh_csum_offset; + uint16_t vrh_bufs; +} __packed; + +#endif /* __NET_BACKENDS_H__ */ diff --git a/usr.sbin/bhyve/pci_e82545.c b/usr.sbin/bhyve/pci_e82545.c index 180590c5282..6e828aefe17 100644 --- a/usr.sbin/bhyve/pci_e82545.c +++ b/usr.sbin/bhyve/pci_e82545.c @@ -66,6 +66,7 @@ __FBSDID("$FreeBSD$"); #include "pci_emul.h" #include "mevent.h" #include "net_utils.h" +#include "net_backends.h" /* Hardware/register definitions XXX: move some to common code. */ #define E82545_VENDOR_ID_INTEL 0x8086 @@ -245,11 +246,10 @@ struct eth_uni { struct e82545_softc { struct pci_devinst *esc_pi; struct vmctx *esc_ctx; - struct mevent *esc_mevp; struct mevent *esc_mevpitr; pthread_mutex_t esc_mtx; struct ether_addr esc_mac; - int esc_tapfd; + net_backend_t *esc_be; /* General */ uint32_t esc_CTRL; /* x0000 device ctl */ @@ -355,7 +355,7 @@ struct e82545_softc { static void e82545_reset(struct e82545_softc *sc, int dev); static void e82545_rx_enable(struct e82545_softc *sc); static void e82545_rx_disable(struct e82545_softc *sc); -static void e82545_tap_callback(int fd, enum ev_type type, void *param); +static void e82545_rx_callback(int fd, enum ev_type type, void *param); static void e82545_tx_start(struct e82545_softc *sc); static void e82545_tx_enable(struct e82545_softc *sc); static void e82545_tx_disable(struct e82545_softc *sc); @@ -824,11 +824,9 @@ e82545_bufsz(uint32_t rctl) return (256); /* Forbidden value. */ } -static uint8_t dummybuf[2048]; - /* XXX one packet at a time until this is debugged */ static void -e82545_tap_callback(int fd, enum ev_type type, void *param) +e82545_rx_callback(int fd, enum ev_type type, void *param) { struct e82545_softc *sc = param; struct e1000_rx_desc *rxd; @@ -843,7 +841,7 @@ e82545_tap_callback(int fd, enum ev_type type, void *param) if (!sc->esc_rx_enabled || sc->esc_rx_loopback) { DPRINTF("rx disabled (!%d || %d) -- packet(s) dropped\r\n", sc->esc_rx_enabled, sc->esc_rx_loopback); - while (read(sc->esc_tapfd, dummybuf, sizeof(dummybuf)) > 0) { + while (netbe_rx_discard(sc->esc_be) > 0) { } goto done1; } @@ -856,7 +854,7 @@ e82545_tap_callback(int fd, enum ev_type type, void *param) if (left < maxpktdesc) { DPRINTF("rx overflow (%d < %d) -- packet(s) dropped\r\n", left, maxpktdesc); - while (read(sc->esc_tapfd, dummybuf, sizeof(dummybuf)) > 0) { + while (netbe_rx_discard(sc->esc_be) > 0) { } goto done1; } @@ -873,9 +871,9 @@ e82545_tap_callback(int fd, enum ev_type type, void *param) rxd->buffer_addr, bufsz); vec[i].iov_len = bufsz; } - len = readv(sc->esc_tapfd, vec, maxpktdesc); + len = netbe_recv(sc->esc_be, vec, maxpktdesc); if (len <= 0) { - DPRINTF("tap: readv() returned %d\n", len); + DPRINTF("netbe_recv() returned %d\n", len); goto done; } @@ -1050,10 +1048,10 @@ static void e82545_transmit_backend(struct e82545_softc *sc, struct iovec *iov, int iovcnt) { - if (sc->esc_tapfd == -1) + if (sc->esc_be == NULL) return; - (void) writev(sc->esc_tapfd, iov, iovcnt); + (void) netbe_send(sc->esc_be, iov, iovcnt); } static void @@ -2209,56 +2207,6 @@ e82545_reset(struct e82545_softc *sc, int drvr) sc->esc_TXDCTL = 0; } -static void -e82545_open_tap(struct e82545_softc *sc, char *opts) -{ - char tbuf[80]; -#ifndef WITHOUT_CAPSICUM - cap_rights_t rights; -#endif - - if (opts == NULL) { - sc->esc_tapfd = -1; - return; - } - - strcpy(tbuf, "/dev/"); - strlcat(tbuf, opts, sizeof(tbuf)); - - sc->esc_tapfd = open(tbuf, O_RDWR); - if (sc->esc_tapfd == -1) { - DPRINTF("unable to open tap device %s\n", opts); - exit(4); - } - - /* - * Set non-blocking and register for read - * notifications with the event loop - */ - int opt = 1; - if (ioctl(sc->esc_tapfd, FIONBIO, &opt) < 0) { - WPRINTF("tap device O_NONBLOCK failed: %d\n", errno); - close(sc->esc_tapfd); - sc->esc_tapfd = -1; - } - -#ifndef WITHOUT_CAPSICUM - cap_rights_init(&rights, CAP_EVENT, CAP_READ, CAP_WRITE); - if (caph_rights_limit(sc->esc_tapfd, &rights) == -1) - errx(EX_OSERR, "Unable to apply rights for sandbox"); -#endif - - sc->esc_mevp = mevent_add(sc->esc_tapfd, - EVF_READ, - e82545_tap_callback, - sc); - if (sc->esc_mevp == NULL) { - DPRINTF("Could not register mevent %d\n", EVF_READ); - close(sc->esc_tapfd); - sc->esc_tapfd = -1; - } -} - static int e82545_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) { @@ -2307,11 +2255,11 @@ e82545_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) E82545_BAR_IO_LEN); /* - * Attempt to open the tap device and read the MAC address + * Attempt to open the net backend and read the MAC address * if specified. Copied from virtio-net, slightly modified. */ mac_provided = 0; - sc->esc_tapfd = -1; + sc->esc_be = NULL; if (opts != NULL) { int err; @@ -2327,11 +2275,10 @@ e82545_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) mac_provided = 1; } - if (strncmp(devname, "tap", 3) == 0 || - strncmp(devname, "vmnet", 5) == 0) - e82545_open_tap(sc, devname); - + err = netbe_init(&sc->esc_be, devname, e82545_rx_callback, sc); free(devname); + if (err) + return (err); } if (!mac_provided) { diff --git a/usr.sbin/bhyve/pci_virtio_net.c b/usr.sbin/bhyve/pci_virtio_net.c index af57a70d025..7799b0aa3ed 100644 --- a/usr.sbin/bhyve/pci_virtio_net.c +++ b/usr.sbin/bhyve/pci_virtio_net.c @@ -32,22 +32,13 @@ __FBSDID("$FreeBSD$"); #include -#ifndef WITHOUT_CAPSICUM -#include -#endif #include #include #include #include #include -#ifndef NETMAP_WITH_LIBS -#define NETMAP_WITH_LIBS -#endif -#include +#include /* IFNAMSIZ */ -#ifndef WITHOUT_CAPSICUM -#include -#endif #include #include #include @@ -58,46 +49,22 @@ __FBSDID("$FreeBSD$"); #include #include #include -#include #include #include -#include #include "bhyverun.h" #include "pci_emul.h" #include "mevent.h" #include "virtio.h" #include "net_utils.h" +#include "net_backends.h" #define VTNET_RINGSZ 1024 #define VTNET_MAXSEGS 256 -/* - * Host capabilities. Note that we only offer a few of these. - */ -#define VIRTIO_NET_F_CSUM (1 << 0) /* host handles partial cksum */ -#define VIRTIO_NET_F_GUEST_CSUM (1 << 1) /* guest handles partial cksum */ -#define VIRTIO_NET_F_MAC (1 << 5) /* host supplies MAC */ -#define VIRTIO_NET_F_GSO_DEPREC (1 << 6) /* deprecated: host handles GSO */ -#define VIRTIO_NET_F_GUEST_TSO4 (1 << 7) /* guest can rcv TSOv4 */ -#define VIRTIO_NET_F_GUEST_TSO6 (1 << 8) /* guest can rcv TSOv6 */ -#define VIRTIO_NET_F_GUEST_ECN (1 << 9) /* guest can rcv TSO with ECN */ -#define VIRTIO_NET_F_GUEST_UFO (1 << 10) /* guest can rcv UFO */ -#define VIRTIO_NET_F_HOST_TSO4 (1 << 11) /* host can rcv TSOv4 */ -#define VIRTIO_NET_F_HOST_TSO6 (1 << 12) /* host can rcv TSOv6 */ -#define VIRTIO_NET_F_HOST_ECN (1 << 13) /* host can rcv TSO with ECN */ -#define VIRTIO_NET_F_HOST_UFO (1 << 14) /* host can rcv UFO */ -#define VIRTIO_NET_F_MRG_RXBUF (1 << 15) /* host can merge RX buffers */ -#define VIRTIO_NET_F_STATUS (1 << 16) /* config status field available */ -#define VIRTIO_NET_F_CTRL_VQ (1 << 17) /* control channel available */ -#define VIRTIO_NET_F_CTRL_RX (1 << 18) /* control channel RX mode support */ -#define VIRTIO_NET_F_CTRL_VLAN (1 << 19) /* control channel VLAN filtering */ -#define VIRTIO_NET_F_GUEST_ANNOUNCE \ - (1 << 21) /* guest can send gratuitous pkts */ - #define VTNET_S_HOSTCAPS \ - ( VIRTIO_NET_F_MAC | VIRTIO_NET_F_MRG_RXBUF | VIRTIO_NET_F_STATUS | \ + ( VIRTIO_NET_F_MAC | VIRTIO_NET_F_STATUS | \ VIRTIO_F_NOTIFY_ON_EMPTY | VIRTIO_RING_F_INDIRECT_DESC) /* @@ -117,19 +84,6 @@ struct virtio_net_config { #define VTNET_MAXQ 3 -/* - * Fixed network header size - */ -struct virtio_net_rxhdr { - uint8_t vrh_flags; - uint8_t vrh_gso_type; - uint16_t vrh_hdr_len; - uint16_t vrh_gso_size; - uint16_t vrh_csum_start; - uint16_t vrh_csum_offset; - uint16_t vrh_bufs; -} __packed; - /* * Debug printf */ @@ -144,20 +98,16 @@ struct pci_vtnet_softc { struct virtio_softc vsc_vs; struct vqueue_info vsc_queues[VTNET_MAXQ - 1]; pthread_mutex_t vsc_mtx; - struct mevent *vsc_mevp; - int vsc_tapfd; - struct nm_desc *vsc_nmd; + net_backend_t *vsc_be; int vsc_rx_ready; int resetting; /* protected by tx_mtx */ uint64_t vsc_features; /* negotiated features */ - struct virtio_net_config vsc_config; - pthread_mutex_t rx_mtx; - int rx_vhdrlen; + unsigned int rx_vhdrlen; int rx_merge; /* merged rx bufs in use */ pthread_t tx_tid; @@ -165,9 +115,8 @@ struct pci_vtnet_softc { pthread_cond_t tx_cond; int tx_in_progress; - void (*pci_vtnet_rx)(struct pci_vtnet_softc *sc); - void (*pci_vtnet_tx)(struct pci_vtnet_softc *sc, struct iovec *iov, - int iovcnt, int len); + struct virtio_net_config vsc_config; + struct virtio_consts vsc_consts; }; static void pci_vtnet_reset(void *); @@ -223,84 +172,20 @@ pci_vtnet_reset(void *vsc) pthread_mutex_unlock(&sc->rx_mtx); } -/* - * Called to send a buffer chain out to the tap device - */ -static void -pci_vtnet_tap_tx(struct pci_vtnet_softc *sc, struct iovec *iov, int iovcnt, - int len) -{ - static char pad[60]; /* all zero bytes */ - - if (sc->vsc_tapfd == -1) - return; - - /* - * If the length is < 60, pad out to that and add the - * extra zero'd segment to the iov. It is guaranteed that - * there is always an extra iov available by the caller. - */ - if (len < 60) { - iov[iovcnt].iov_base = pad; - iov[iovcnt].iov_len = 60 - len; - iovcnt++; - } - (void) writev(sc->vsc_tapfd, iov, iovcnt); -} - -/* - * Called when there is read activity on the tap file descriptor. - * Each buffer posted by the guest is assumed to be able to contain - * an entire ethernet frame + rx header. - * MP note: the dummybuf is only used for discarding frames, so there - * is no need for it to be per-vtnet or locked. - */ -static uint8_t dummybuf[2048]; - -static __inline struct iovec * -rx_iov_trim(struct iovec *iov, int *niov, int tlen) -{ - struct iovec *riov; - - /* XXX short-cut: assume first segment is >= tlen */ - assert(iov[0].iov_len >= tlen); - - iov[0].iov_len -= tlen; - if (iov[0].iov_len == 0) { - assert(*niov > 1); - *niov -= 1; - riov = &iov[1]; - } else { - iov[0].iov_base = (void *)((uintptr_t)iov[0].iov_base + tlen); - riov = &iov[0]; - } - - return (riov); -} - static void -pci_vtnet_tap_rx(struct pci_vtnet_softc *sc) +pci_vtnet_rx(struct pci_vtnet_softc *sc) { - struct iovec iov[VTNET_MAXSEGS], *riov; + struct iovec iov[VTNET_MAXSEGS + 1]; struct vqueue_info *vq; - void *vrx; int len, n; uint16_t idx; - /* - * Should never be called without a valid tap fd - */ - assert(sc->vsc_tapfd != -1); - - /* - * But, will be called when the rx ring hasn't yet - * been set up. - */ if (!sc->vsc_rx_ready) { /* + * The rx ring has not yet been set up. * Drop the packet and try later. */ - (void) read(sc->vsc_tapfd, dummybuf, sizeof(dummybuf)); + netbe_rx_discard(sc->vsc_be); return; } @@ -310,11 +195,11 @@ pci_vtnet_tap_rx(struct pci_vtnet_softc *sc) vq = &sc->vsc_queues[VTNET_RXQ]; if (!vq_has_descs(vq)) { /* - * Drop the packet and try later. Interrupt on - * empty, if that's negotiated. + * No available rx buffers. Drop the packet and try later. + * Interrupt on empty, if that's negotiated. */ - (void) read(sc->vsc_tapfd, dummybuf, sizeof(dummybuf)); - vq_endchains(vq, 1); + netbe_rx_discard(sc->vsc_be); + vq_endchains(vq, /*used_all_avail=*/1); return; } @@ -325,256 +210,44 @@ pci_vtnet_tap_rx(struct pci_vtnet_softc *sc) n = vq_getchain(vq, &idx, iov, VTNET_MAXSEGS, NULL); assert(n >= 1 && n <= VTNET_MAXSEGS); - /* - * Get a pointer to the rx header, and use the - * data immediately following it for the packet buffer. - */ - vrx = iov[0].iov_base; - riov = rx_iov_trim(iov, &n, sc->rx_vhdrlen); + len = netbe_recv(sc->vsc_be, iov, n); - len = readv(sc->vsc_tapfd, riov, n); - - if (len < 0 && errno == EWOULDBLOCK) { + if (len <= 0) { /* - * No more packets, but still some avail ring - * entries. Interrupt if needed/appropriate. + * No more packets (len == 0), or backend errored + * (err < 0). Return unused available buffers. */ vq_retchain(vq); - vq_endchains(vq, 0); + /* Interrupt if needed/appropriate and stop. */ + vq_endchains(vq, /*used_all_avail=*/0); return; } - /* - * The only valid field in the rx packet header is the - * number of buffers if merged rx bufs were negotiated. - */ - memset(vrx, 0, sc->rx_vhdrlen); - - if (sc->rx_merge) { - struct virtio_net_rxhdr *vrxh; - - vrxh = vrx; - vrxh->vrh_bufs = 1; - } - - /* - * Release this chain and handle more chains. - */ - vq_relchain(vq, idx, len + sc->rx_vhdrlen); + /* Publish the info to the guest */ + vq_relchain(vq, idx, (uint32_t)len); } while (vq_has_descs(vq)); /* Interrupt if needed, including for NOTIFY_ON_EMPTY. */ - vq_endchains(vq, 1); -} - -static __inline int -pci_vtnet_netmap_writev(struct nm_desc *nmd, struct iovec *iov, int iovcnt) -{ - int r, i; - int len = 0; - - for (r = nmd->cur_tx_ring; ; ) { - struct netmap_ring *ring = NETMAP_TXRING(nmd->nifp, r); - uint32_t cur, idx; - char *buf; - - if (nm_ring_empty(ring)) { - r++; - if (r > nmd->last_tx_ring) - r = nmd->first_tx_ring; - if (r == nmd->cur_tx_ring) - break; - continue; - } - cur = ring->cur; - idx = ring->slot[cur].buf_idx; - buf = NETMAP_BUF(ring, idx); - - for (i = 0; i < iovcnt; i++) { - if (len + iov[i].iov_len > 2048) - break; - memcpy(&buf[len], iov[i].iov_base, iov[i].iov_len); - len += iov[i].iov_len; - } - ring->slot[cur].len = len; - ring->head = ring->cur = nm_ring_next(ring, cur); - nmd->cur_tx_ring = r; - ioctl(nmd->fd, NIOCTXSYNC, NULL); - break; - } - - return (len); -} - -static __inline int -pci_vtnet_netmap_readv(struct nm_desc *nmd, struct iovec *iov, int iovcnt) -{ - int len = 0; - int i = 0; - int r; - - for (r = nmd->cur_rx_ring; ; ) { - struct netmap_ring *ring = NETMAP_RXRING(nmd->nifp, r); - uint32_t cur, idx; - char *buf; - size_t left; - - if (nm_ring_empty(ring)) { - r++; - if (r > nmd->last_rx_ring) - r = nmd->first_rx_ring; - if (r == nmd->cur_rx_ring) - break; - continue; - } - cur = ring->cur; - idx = ring->slot[cur].buf_idx; - buf = NETMAP_BUF(ring, idx); - left = ring->slot[cur].len; - - for (i = 0; i < iovcnt && left > 0; i++) { - if (iov[i].iov_len > left) - iov[i].iov_len = left; - memcpy(iov[i].iov_base, &buf[len], iov[i].iov_len); - len += iov[i].iov_len; - left -= iov[i].iov_len; - } - ring->head = ring->cur = nm_ring_next(ring, cur); - nmd->cur_rx_ring = r; - ioctl(nmd->fd, NIOCRXSYNC, NULL); - break; - } - for (; i < iovcnt; i++) - iov[i].iov_len = 0; - - return (len); + vq_endchains(vq, /*used_all_avail=*/1); } /* - * Called to send a buffer chain out to the vale port + * Called when there is read activity on the backend file descriptor. + * Each buffer posted by the guest is assumed to be able to contain + * an entire ethernet frame + rx header. */ -static void -pci_vtnet_netmap_tx(struct pci_vtnet_softc *sc, struct iovec *iov, int iovcnt, - int len) -{ - static char pad[60]; /* all zero bytes */ - - if (sc->vsc_nmd == NULL) - return; - - /* - * If the length is < 60, pad out to that and add the - * extra zero'd segment to the iov. It is guaranteed that - * there is always an extra iov available by the caller. - */ - if (len < 60) { - iov[iovcnt].iov_base = pad; - iov[iovcnt].iov_len = 60 - len; - iovcnt++; - } - (void) pci_vtnet_netmap_writev(sc->vsc_nmd, iov, iovcnt); -} - -static void -pci_vtnet_netmap_rx(struct pci_vtnet_softc *sc) -{ - struct iovec iov[VTNET_MAXSEGS], *riov; - struct vqueue_info *vq; - void *vrx; - int len, n; - uint16_t idx; - - /* - * Should never be called without a valid netmap descriptor - */ - assert(sc->vsc_nmd != NULL); - - /* - * But, will be called when the rx ring hasn't yet - * been set up. - */ - if (!sc->vsc_rx_ready) { - /* - * Drop the packet and try later. - */ - (void) nm_nextpkt(sc->vsc_nmd, (void *)dummybuf); - return; - } - - /* - * Check for available rx buffers - */ - vq = &sc->vsc_queues[VTNET_RXQ]; - if (!vq_has_descs(vq)) { - /* - * Drop the packet and try later. Interrupt on - * empty, if that's negotiated. - */ - (void) nm_nextpkt(sc->vsc_nmd, (void *)dummybuf); - vq_endchains(vq, 1); - return; - } - - do { - /* - * Get descriptor chain. - */ - n = vq_getchain(vq, &idx, iov, VTNET_MAXSEGS, NULL); - assert(n >= 1 && n <= VTNET_MAXSEGS); - - /* - * Get a pointer to the rx header, and use the - * data immediately following it for the packet buffer. - */ - vrx = iov[0].iov_base; - riov = rx_iov_trim(iov, &n, sc->rx_vhdrlen); - - len = pci_vtnet_netmap_readv(sc->vsc_nmd, riov, n); - - if (len == 0) { - /* - * No more packets, but still some avail ring - * entries. Interrupt if needed/appropriate. - */ - vq_retchain(vq); - vq_endchains(vq, 0); - return; - } - - /* - * The only valid field in the rx packet header is the - * number of buffers if merged rx bufs were negotiated. - */ - memset(vrx, 0, sc->rx_vhdrlen); - - if (sc->rx_merge) { - struct virtio_net_rxhdr *vrxh; - - vrxh = vrx; - vrxh->vrh_bufs = 1; - } - - /* - * Release this chain and handle more chains. - */ - vq_relchain(vq, idx, len + sc->rx_vhdrlen); - } while (vq_has_descs(vq)); - - /* Interrupt if needed, including for NOTIFY_ON_EMPTY. */ - vq_endchains(vq, 1); -} - static void pci_vtnet_rx_callback(int fd, enum ev_type type, void *param) { struct pci_vtnet_softc *sc = param; pthread_mutex_lock(&sc->rx_mtx); - sc->pci_vtnet_rx(sc); + pci_vtnet_rx(sc); pthread_mutex_unlock(&sc->rx_mtx); } +/* Called on RX kick. */ static void pci_vtnet_ping_rxq(void *vsc, struct vqueue_info *vq) { @@ -591,35 +264,29 @@ pci_vtnet_ping_rxq(void *vsc, struct vqueue_info *vq) pthread_mutex_unlock(&sc->rx_mtx); } +/* TX virtqueue processing, called by the TX thread. */ static void pci_vtnet_proctx(struct pci_vtnet_softc *sc, struct vqueue_info *vq) { struct iovec iov[VTNET_MAXSEGS + 1]; - int i, n; - int plen, tlen; uint16_t idx; + ssize_t len; + int n; /* - * Obtain chain of descriptors. The first one is - * really the header descriptor, so we need to sum - * up two lengths: packet length and transfer length. + * Obtain chain of descriptors. The first descriptor also + * contains the virtio-net header. */ n = vq_getchain(vq, &idx, iov, VTNET_MAXSEGS, NULL); assert(n >= 1 && n <= VTNET_MAXSEGS); - plen = 0; - tlen = iov[0].iov_len; - for (i = 1; i < n; i++) { - plen += iov[i].iov_len; - tlen += iov[i].iov_len; - } - DPRINTF(("virtio: packet send, %d bytes, %d segs\n\r", plen, n)); - sc->pci_vtnet_tx(sc, &iov[1], n - 1, plen); + len = netbe_send(sc->vsc_be, iov, n); - /* chain is processed, release it and set tlen */ - vq_relchain(vq, idx, tlen); + /* chain is processed, release it and set len */ + vq_relchain(vq, idx, len > 0 ? len : 0); } +/* Called on TX kick. */ static void pci_vtnet_ping_txq(void *vsc, struct vqueue_info *vq) { @@ -686,7 +353,7 @@ pci_vtnet_tx_thread(void *param) /* * Generate an interrupt if needed. */ - vq_endchains(vq, 1); + vq_endchains(vq, /*used_all_avail=*/1); pthread_mutex_lock(&sc->tx_mtx); } @@ -701,93 +368,23 @@ pci_vtnet_ping_ctlq(void *vsc, struct vqueue_info *vq) } #endif -static void -pci_vtnet_tap_setup(struct pci_vtnet_softc *sc, char *devname) -{ - char tbuf[80]; -#ifndef WITHOUT_CAPSICUM - cap_rights_t rights; -#endif - - strcpy(tbuf, "/dev/"); - strlcat(tbuf, devname, sizeof(tbuf)); - - sc->pci_vtnet_rx = pci_vtnet_tap_rx; - sc->pci_vtnet_tx = pci_vtnet_tap_tx; - - sc->vsc_tapfd = open(tbuf, O_RDWR); - if (sc->vsc_tapfd == -1) { - WPRINTF(("open of tap device %s failed\n", tbuf)); - return; - } - - /* - * Set non-blocking and register for read - * notifications with the event loop - */ - int opt = 1; - if (ioctl(sc->vsc_tapfd, FIONBIO, &opt) < 0) { - WPRINTF(("tap device O_NONBLOCK failed\n")); - close(sc->vsc_tapfd); - sc->vsc_tapfd = -1; - } - -#ifndef WITHOUT_CAPSICUM - cap_rights_init(&rights, CAP_EVENT, CAP_READ, CAP_WRITE); - if (caph_rights_limit(sc->vsc_tapfd, &rights) == -1) - errx(EX_OSERR, "Unable to apply rights for sandbox"); -#endif - - sc->vsc_mevp = mevent_add(sc->vsc_tapfd, - EVF_READ, - pci_vtnet_rx_callback, - sc); - if (sc->vsc_mevp == NULL) { - WPRINTF(("Could not register event\n")); - close(sc->vsc_tapfd); - sc->vsc_tapfd = -1; - } -} - -static void -pci_vtnet_netmap_setup(struct pci_vtnet_softc *sc, char *ifname) -{ - sc->pci_vtnet_rx = pci_vtnet_netmap_rx; - sc->pci_vtnet_tx = pci_vtnet_netmap_tx; - - sc->vsc_nmd = nm_open(ifname, NULL, 0, 0); - if (sc->vsc_nmd == NULL) { - WPRINTF(("open of netmap device %s failed\n", ifname)); - return; - } - - sc->vsc_mevp = mevent_add(sc->vsc_nmd->fd, - EVF_READ, - pci_vtnet_rx_callback, - sc); - if (sc->vsc_mevp == NULL) { - WPRINTF(("Could not register event\n")); - nm_close(sc->vsc_nmd); - sc->vsc_nmd = NULL; - } -} - static int pci_vtnet_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) { - char tname[MAXCOMLEN + 1]; struct pci_vtnet_softc *sc; - char *devname; - char *vtopts; + char tname[MAXCOMLEN + 1]; int mac_provided; + /* + * Allocate data structures for further virtio initializations. + * sc also contains a copy of vtnet_vi_consts, since capabilities + * change depending on the backend. + */ sc = calloc(1, sizeof(struct pci_vtnet_softc)); + sc->vsc_consts = vtnet_vi_consts; pthread_mutex_init(&sc->vsc_mtx, NULL); - vi_softc_linkup(&sc->vsc_vs, &vtnet_vi_consts, sc, pi, sc->vsc_queues); - sc->vsc_vs.vs_mtx = &sc->vsc_mtx; - sc->vsc_queues[VTNET_RXQ].vq_qsize = VTNET_RINGSZ; sc->vsc_queues[VTNET_RXQ].vq_notify = pci_vtnet_ping_rxq; sc->vsc_queues[VTNET_TXQ].vq_qsize = VTNET_RINGSZ; @@ -798,13 +395,13 @@ pci_vtnet_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) #endif /* - * Attempt to open the tap device and read the MAC address - * if specified + * Attempt to open the backend device and read the MAC address + * if specified. */ mac_provided = 0; - sc->vsc_tapfd = -1; - sc->vsc_nmd = NULL; if (opts != NULL) { + char *devname; + char *vtopts; int err; devname = vtopts = strdup(opts); @@ -819,13 +416,12 @@ pci_vtnet_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) mac_provided = 1; } - if (strncmp(devname, "vale", 4) == 0) - pci_vtnet_netmap_setup(sc, devname); - if (strncmp(devname, "tap", 3) == 0 || - strncmp(devname, "vmnet", 5) == 0) - pci_vtnet_tap_setup(sc, devname); - + err = netbe_init(&sc->vsc_be, devname, pci_vtnet_rx_callback, + sc); free(devname); + if (err) + return (err); + sc->vsc_consts.vc_hv_caps |= netbe_get_cap(sc->vsc_be); } if (!mac_provided) { @@ -839,10 +435,12 @@ pci_vtnet_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) pci_set_cfgdata16(pi, PCIR_SUBDEV_0, VIRTIO_TYPE_NET); pci_set_cfgdata16(pi, PCIR_SUBVEND_0, VIRTIO_VENDOR); - /* Link is up if we managed to open tap device or vale port. */ - sc->vsc_config.status = (opts == NULL || sc->vsc_tapfd >= 0 || - sc->vsc_nmd != NULL); + /* Link is up if we managed to open backend device. */ + sc->vsc_config.status = (opts == NULL || sc->vsc_be); + vi_softc_linkup(&sc->vsc_vs, &sc->vsc_consts, sc, pi, sc->vsc_queues); + sc->vsc_vs.vs_mtx = &sc->vsc_mtx; + /* use BAR 1 to map MSI-X table and PBA, if we're using MSI-X */ if (vi_intr_init(&sc->vsc_vs, 1, fbsdrun_virtio_msix())) return (1); @@ -878,8 +476,8 @@ pci_vtnet_cfgwrite(void *vsc, int offset, int size, uint32_t value) struct pci_vtnet_softc *sc = vsc; void *ptr; - if (offset < 6) { - assert(offset + size <= 6); + if (offset < (int)sizeof(sc->vsc_config.mac)) { + assert(offset + size <= (int)sizeof(sc->vsc_config.mac)); /* * The driver is allowed to change the MAC address */ @@ -911,14 +509,18 @@ pci_vtnet_neg_features(void *vsc, uint64_t negotiated_features) sc->vsc_features = negotiated_features; - if (!(sc->vsc_features & VIRTIO_NET_F_MRG_RXBUF)) { + if (!(negotiated_features & VIRTIO_NET_F_MRG_RXBUF)) { sc->rx_merge = 0; - /* non-merge rx header is 2 bytes shorter */ - sc->rx_vhdrlen -= 2; + /* Without mergeable rx buffers, virtio-net header is 2 + * bytes shorter than sizeof(struct virtio_net_rxhdr). */ + sc->rx_vhdrlen = sizeof(struct virtio_net_rxhdr) - 2; } + + /* Tell the backend to enable some capabilities it has advertised. */ + netbe_set_cap(sc->vsc_be, negotiated_features, sc->rx_vhdrlen); } -struct pci_devemu pci_de_vnet = { +static struct pci_devemu pci_de_vnet = { .pe_emu = "virtio-net", .pe_init = pci_vtnet_init, .pe_barwrite = vi_pci_write, -- 2.45.0