Author: grehan
Date: Tue Sep 16 19:08:54 2014
New Revision: 271685
URL: http://svnweb.freebsd.org/changeset/base/271685
Log:
MFC virtio-net changes.
Re-tested with NetBSD/amd64 5.2.2, 6.1.4 and 7-beta.
r271299:
Add a callback to be notified about negotiated features.
r271338:
Allow vtnet operation without merged rx buffers.
NetBSD's virtio-net implementation doesn't negotiate
the merged rx-buffers feature. To support this, check
to see if the feature was negotiated, and then adjust
the operation of the receive path accordingly by using
a larger iovec, and a smaller rx header.
In addition, ignore writes to the (read-only) status byte.
Approved by: re (glebius)
Obtained from: Vincenzo Maffione, Universita` di Pisa (r271299)
Modified:
stable/10/usr.sbin/bhyve/pci_virtio_block.c
stable/10/usr.sbin/bhyve/pci_virtio_net.c
stable/10/usr.sbin/bhyve/pci_virtio_rnd.c
stable/10/usr.sbin/bhyve/virtio.c
stable/10/usr.sbin/bhyve/virtio.h
Directory Properties:
stable/10/ (props changed)
Modified: stable/10/usr.sbin/bhyve/pci_virtio_block.c
==============================================================================
--- stable/10/usr.sbin/bhyve/pci_virtio_block.c Tue Sep 16 18:02:24 2014
(r271684)
+++ stable/10/usr.sbin/bhyve/pci_virtio_block.c Tue Sep 16 19:08:54 2014
(r271685)
@@ -133,6 +133,7 @@ static struct virtio_consts vtblk_vi_con
pci_vtblk_notify, /* device-wide qnotify */
pci_vtblk_cfgread, /* read PCI config */
pci_vtblk_cfgwrite, /* write PCI config */
+ NULL, /* apply negotiated features */
VTBLK_S_HOSTCAPS, /* our capabilities */
};
Modified: stable/10/usr.sbin/bhyve/pci_virtio_net.c
==============================================================================
--- stable/10/usr.sbin/bhyve/pci_virtio_net.c Tue Sep 16 18:02:24 2014
(r271684)
+++ stable/10/usr.sbin/bhyve/pci_virtio_net.c Tue Sep 16 19:08:54 2014
(r271685)
@@ -135,11 +135,14 @@ struct pci_vtnet_softc {
int vsc_rx_ready;
volatile int resetting; /* set and checked outside lock */
- uint32_t vsc_features;
+ uint64_t vsc_features; /* negotiated features */
+
struct virtio_net_config vsc_config;
pthread_mutex_t rx_mtx;
int rx_in_progress;
+ int rx_vhdrlen;
+ int rx_merge; /* merged rx bufs in use */
pthread_t tx_tid;
pthread_mutex_t tx_mtx;
@@ -151,6 +154,7 @@ static void pci_vtnet_reset(void *);
/* static void pci_vtnet_notify(void *, struct vqueue_info *); */
static int pci_vtnet_cfgread(void *, int, int, uint32_t *);
static int pci_vtnet_cfgwrite(void *, int, int, uint32_t);
+static void pci_vtnet_neg_features(void *, uint64_t);
static struct virtio_consts vtnet_vi_consts = {
"vtnet", /* our name */
@@ -160,6 +164,7 @@ static struct virtio_consts vtnet_vi_con
NULL, /* device-wide qnotify -- not used */
pci_vtnet_cfgread, /* read PCI config */
pci_vtnet_cfgwrite, /* write PCI config */
+ pci_vtnet_neg_features, /* apply negotiated features */
VTNET_S_HOSTCAPS, /* our capabilities */
};
@@ -212,6 +217,8 @@ pci_vtnet_reset(void *vsc)
pci_vtnet_rxwait(sc);
sc->vsc_rx_ready = 0;
+ sc->rx_merge = 1;
+ sc->rx_vhdrlen = sizeof(struct virtio_net_rxhdr);
/* now reset rings, MSI-X vectors, and negotiated capabilities */
vi_reset_dev(&sc->vsc_vs);
@@ -253,14 +260,34 @@ pci_vtnet_tap_tx(struct pci_vtnet_softc
*/
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)
{
+ struct iovec iov[VTNET_MAXSEGS], *riov;
struct vqueue_info *vq;
- struct virtio_net_rxhdr *vrx;
- uint8_t *buf;
- int len;
- struct iovec iov;
+ void *vrx;
+ int len, n;
/*
* Should never be called without a valid tap fd
@@ -296,21 +323,19 @@ pci_vtnet_tap_rx(struct pci_vtnet_softc
do {
/*
- * Get descriptor chain, which should have just
- * one descriptor in it.
- * ??? allow guests to use multiple descs?
+ * Get descriptor chain.
*/
- assert(vq_getchain(vq, &iov, 1, NULL) == 1);
+ n = vq_getchain(vq, 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.iov_base;
- buf = (uint8_t *)(vrx + 1);
+ vrx = iov[0].iov_base;
+ riov = rx_iov_trim(iov, &n, sc->rx_vhdrlen);
- len = read(sc->vsc_tapfd, buf,
- iov.iov_len - sizeof(struct virtio_net_rxhdr));
+ len = readv(sc->vsc_tapfd, riov, n);
if (len < 0 && errno == EWOULDBLOCK) {
/*
@@ -323,16 +348,21 @@ pci_vtnet_tap_rx(struct pci_vtnet_softc
/*
* The only valid field in the rx packet header is the
- * number of buffers, which is always 1 without TSO
- * support.
+ * number of buffers if merged rx bufs were negotiated.
*/
- memset(vrx, 0, sizeof(struct virtio_net_rxhdr));
- vrx->vrh_bufs = 1;
+ 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, len + sizeof(struct virtio_net_rxhdr));
+ vq_relchain(vq, len + sc->rx_vhdrlen);
} while (vq_has_descs(vq));
/* Interrupt if needed, including for NOTIFY_ON_EMPTY. */
@@ -623,6 +653,8 @@ pci_vtnet_init(struct vmctx *ctx, struct
sc->resetting = 0;
+ sc->rx_merge = 1;
+ sc->rx_vhdrlen = sizeof(struct virtio_net_rxhdr);
sc->rx_in_progress = 0;
pthread_mutex_init(&sc->rx_mtx, NULL);
@@ -656,9 +688,10 @@ pci_vtnet_cfgwrite(void *vsc, int offset
ptr = &sc->vsc_config.mac[offset];
memcpy(ptr, &value, size);
} else {
+ /* silently ignore other writes */
DPRINTF(("vtnet: write to readonly reg %d\n\r", offset));
- return (1);
}
+
return (0);
}
@@ -673,6 +706,20 @@ pci_vtnet_cfgread(void *vsc, int offset,
return (0);
}
+static void
+pci_vtnet_neg_features(void *vsc, uint64_t negotiated_features)
+{
+ struct pci_vtnet_softc *sc = vsc;
+
+ sc->vsc_features = negotiated_features;
+
+ if (!(sc->vsc_features & VIRTIO_NET_F_MRG_RXBUF)) {
+ sc->rx_merge = 0;
+ /* non-merge rx header is 2 bytes shorter */
+ sc->rx_vhdrlen -= 2;
+ }
+}
+
struct pci_devemu pci_de_vnet = {
.pe_emu = "virtio-net",
.pe_init = pci_vtnet_init,
Modified: stable/10/usr.sbin/bhyve/pci_virtio_rnd.c
==============================================================================
--- stable/10/usr.sbin/bhyve/pci_virtio_rnd.c Tue Sep 16 18:02:24 2014
(r271684)
+++ stable/10/usr.sbin/bhyve/pci_virtio_rnd.c Tue Sep 16 19:08:54 2014
(r271685)
@@ -80,6 +80,7 @@ static struct virtio_consts vtrnd_vi_con
pci_vtrnd_notify, /* device-wide qnotify */
NULL, /* read virtio config */
NULL, /* write virtio config */
+ NULL, /* apply negotiated features */
0, /* our capabilities */
};
Modified: stable/10/usr.sbin/bhyve/virtio.c
==============================================================================
--- stable/10/usr.sbin/bhyve/virtio.c Tue Sep 16 18:02:24 2014
(r271684)
+++ stable/10/usr.sbin/bhyve/virtio.c Tue Sep 16 19:08:54 2014
(r271685)
@@ -694,6 +694,9 @@ bad:
switch (offset) {
case VTCFG_R_GUESTCAP:
vs->vs_negotiated_caps = value & vc->vc_hv_caps;
+ if (vc->vc_apply_features)
+ (*vc->vc_apply_features)(DEV_SOFTC(vs),
+ vs->vs_negotiated_caps);
break;
case VTCFG_R_PFN:
if (vs->vs_curq >= vc->vc_nvq)
Modified: stable/10/usr.sbin/bhyve/virtio.h
==============================================================================
--- stable/10/usr.sbin/bhyve/virtio.h Tue Sep 16 18:02:24 2014
(r271684)
+++ stable/10/usr.sbin/bhyve/virtio.h Tue Sep 16 19:08:54 2014
(r271685)
@@ -352,6 +352,8 @@ struct virtio_consts {
/* called to read config regs */
int (*vc_cfgwrite)(void *, int, int, uint32_t);
/* called to write config regs */
+ void (*vc_apply_features)(void *, uint64_t);
+ /* called to apply negotiated features */
uint64_t vc_hv_caps; /* hypervisor-provided capabilities */
};
_______________________________________________
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
