[qemu.git] / hw / xen_nic.c

/*
 *  xen paravirt network card backend
 *
 *  (c) Gerd Hoffmann <kraxel@redhat.com>
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; under version 2 of the License.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License along
 *  with this program; if not, see <http://www.gnu.org/licenses/>.
 */

#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <string.h>
#include <unistd.h>
#include <signal.h>
#include <inttypes.h>
#include <fcntl.h>
#include <errno.h>
#include <pthread.h>
#include <sys/socket.h>
#include <sys/ioctl.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mman.h>
#include <sys/wait.h>

#include <xs.h>
#include <xenctrl.h>
#include <xen/io/xenbus.h>
#include <xen/io/netif.h>

#include "hw.h"
#include "net.h"
#include "net/checksum.h"
#include "net/util.h"
#include "qemu-char.h"
#include "xen_backend.h"

/* ------------------------------------------------------------- */

struct XenNetDev {
    struct XenDevice      xendev;  /* must be first */
    char                  *mac;
    int                   tx_work;
    int                   tx_ring_ref;
    int                   rx_ring_ref;
    struct netif_tx_sring *txs;
    struct netif_rx_sring *rxs;
    netif_tx_back_ring_t  tx_ring;
    netif_rx_back_ring_t  rx_ring;
    NICConf               conf;
    NICState              *nic;
};

/* ------------------------------------------------------------- */

static void net_tx_response(struct XenNetDev *netdev, netif_tx_request_t *txp, int8_t st)
{
    RING_IDX i = netdev->tx_ring.rsp_prod_pvt;
    netif_tx_response_t *resp;
    int notify;

    resp = RING_GET_RESPONSE(&netdev->tx_ring, i);
    resp->id     = txp->id;
    resp->status = st;

#if 0
    if (txp->flags & NETTXF_extra_info) {
        RING_GET_RESPONSE(&netdev->tx_ring, ++i)->status = NETIF_RSP_NULL;
    }
#endif

    netdev->tx_ring.rsp_prod_pvt = ++i;
    RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->tx_ring, notify);
    if (notify) {
        xen_be_send_notify(&netdev->xendev);
    }

    if (i == netdev->tx_ring.req_cons) {
        int more_to_do;
        RING_FINAL_CHECK_FOR_REQUESTS(&netdev->tx_ring, more_to_do);
        if (more_to_do) {
            netdev->tx_work++;
        }
    }
}

static void net_tx_error(struct XenNetDev *netdev, netif_tx_request_t *txp, RING_IDX end)
{
#if 0
    /*
     * Hmm, why netback fails everything in the ring?
     * Should we do that even when not supporting SG and TSO?
     */
    RING_IDX cons = netdev->tx_ring.req_cons;

    do {
        make_tx_response(netif, txp, NETIF_RSP_ERROR);
        if (cons >= end) {
            break;
        }
        txp = RING_GET_REQUEST(&netdev->tx_ring, cons++);
    } while (1);
    netdev->tx_ring.req_cons = cons;
    netif_schedule_work(netif);
    netif_put(netif);
#else
    net_tx_response(netdev, txp, NETIF_RSP_ERROR);
#endif
}

static void net_tx_packets(struct XenNetDev *netdev)
{
    netif_tx_request_t txreq;
    RING_IDX rc, rp;
    void *page;
    void *tmpbuf = NULL;

    for (;;) {
        rc = netdev->tx_ring.req_cons;
        rp = netdev->tx_ring.sring->req_prod;
        xen_rmb(); /* Ensure we see queued requests up to 'rp'. */

        while ((rc != rp)) {
            if (RING_REQUEST_CONS_OVERFLOW(&netdev->tx_ring, rc)) {
                break;
            }
            memcpy(&txreq, RING_GET_REQUEST(&netdev->tx_ring, rc), sizeof(txreq));
            netdev->tx_ring.req_cons = ++rc;

#if 1
            /* should not happen in theory, we don't announce the *
             * feature-{sg,gso,whatelse} flags in xenstore (yet?) */
            if (txreq.flags & NETTXF_extra_info) {
                xen_be_printf(&netdev->xendev, 0, "FIXME: extra info flag\n");
                net_tx_error(netdev, &txreq, rc);
                continue;
            }
            if (txreq.flags & NETTXF_more_data) {
                xen_be_printf(&netdev->xendev, 0, "FIXME: more data flag\n");
                net_tx_error(netdev, &txreq, rc);
                continue;
            }
#endif

            if (txreq.size < 14) {
                xen_be_printf(&netdev->xendev, 0, "bad packet size: %d\n", txreq.size);
                net_tx_error(netdev, &txreq, rc);
                continue;
            }

            if ((txreq.offset + txreq.size) > XC_PAGE_SIZE) {
                xen_be_printf(&netdev->xendev, 0, "error: page crossing\n");
                net_tx_error(netdev, &txreq, rc);
                continue;
            }

            xen_be_printf(&netdev->xendev, 3, "tx packet ref %d, off %d, len %d, flags 0x%x%s%s%s%s\n",
                          txreq.gref, txreq.offset, txreq.size, txreq.flags,
                          (txreq.flags & NETTXF_csum_blank)     ? " csum_blank"     : "",
                          (txreq.flags & NETTXF_data_validated) ? " data_validated" : "",
                          (txreq.flags & NETTXF_more_data)      ? " more_data"      : "",
                          (txreq.flags & NETTXF_extra_info)     ? " extra_info"     : "");

            page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
                                           netdev->xendev.dom,
                                           txreq.gref, PROT_READ);
            if (page == NULL) {
                xen_be_printf(&netdev->xendev, 0, "error: tx gref dereference failed (%d)\n",
                              txreq.gref);
                net_tx_error(netdev, &txreq, rc);
                continue;
            }
            if (txreq.flags & NETTXF_csum_blank) {
                /* have read-only mapping -> can't fill checksum in-place */
                if (!tmpbuf) {
                    tmpbuf = qemu_malloc(XC_PAGE_SIZE);
                }
                memcpy(tmpbuf, page + txreq.offset, txreq.size);
                net_checksum_calculate(tmpbuf, txreq.size);
                qemu_send_packet(&netdev->nic->nc, tmpbuf, txreq.size);
            } else {
                qemu_send_packet(&netdev->nic->nc, page + txreq.offset, txreq.size);
            }
            xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
            net_tx_response(netdev, &txreq, NETIF_RSP_OKAY);
        }
        if (!netdev->tx_work) {
            break;
        }
        netdev->tx_work = 0;
    }
    qemu_free(tmpbuf);
}

/* ------------------------------------------------------------- */

static void net_rx_response(struct XenNetDev *netdev,
                            netif_rx_request_t *req, int8_t st,
                            uint16_t offset, uint16_t size,
                            uint16_t flags)
{
    RING_IDX i = netdev->rx_ring.rsp_prod_pvt;
    netif_rx_response_t *resp;
    int notify;

    resp = RING_GET_RESPONSE(&netdev->rx_ring, i);
    resp->offset     = offset;
    resp->flags      = flags;
    resp->id         = req->id;
    resp->status     = (int16_t)size;
    if (st < 0) {
        resp->status = (int16_t)st;
    }

    xen_be_printf(&netdev->xendev, 3, "rx response: idx %d, status %d, flags 0x%x\n",
                  i, resp->status, resp->flags);

    netdev->rx_ring.rsp_prod_pvt = ++i;
    RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->rx_ring, notify);
    if (notify) {
        xen_be_send_notify(&netdev->xendev);
    }
}

#define NET_IP_ALIGN 2

static int net_rx_ok(VLANClientState *nc)
{
    struct XenNetDev *netdev = DO_UPCAST(NICState, nc, nc)->opaque;
    RING_IDX rc, rp;

    if (netdev->xendev.be_state != XenbusStateConnected) {
        return 0;
    }

    rc = netdev->rx_ring.req_cons;
    rp = netdev->rx_ring.sring->req_prod;
    xen_rmb();

    if (rc == rp || RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
        xen_be_printf(&netdev->xendev, 2, "%s: no rx buffers (%d/%d)\n",
                      __FUNCTION__, rc, rp);
        return 0;
    }
    return 1;
}

static ssize_t net_rx_packet(VLANClientState *nc, const uint8_t *buf, size_t size)
{
    struct XenNetDev *netdev = DO_UPCAST(NICState, nc, nc)->opaque;
    netif_rx_request_t rxreq;
    RING_IDX rc, rp;
    void *page;

    if (netdev->xendev.be_state != XenbusStateConnected) {
        return -1;
    }

    rc = netdev->rx_ring.req_cons;
    rp = netdev->rx_ring.sring->req_prod;
    xen_rmb(); /* Ensure we see queued requests up to 'rp'. */

    if (rc == rp || RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
        xen_be_printf(&netdev->xendev, 2, "no buffer, drop packet\n");
        return -1;
    }
    if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
        xen_be_printf(&netdev->xendev, 0, "packet too big (%lu > %ld)",
                      (unsigned long)size, XC_PAGE_SIZE - NET_IP_ALIGN);
        return -1;
    }

    memcpy(&rxreq, RING_GET_REQUEST(&netdev->rx_ring, rc), sizeof(rxreq));
    netdev->rx_ring.req_cons = ++rc;

    page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
                                   netdev->xendev.dom,
                                   rxreq.gref, PROT_WRITE);
    if (page == NULL) {
        xen_be_printf(&netdev->xendev, 0, "error: rx gref dereference failed (%d)\n",
                      rxreq.gref);
        net_rx_response(netdev, &rxreq, NETIF_RSP_ERROR, 0, 0, 0);
        return -1;
    }
    memcpy(page + NET_IP_ALIGN, buf, size);
    xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
    net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);

    return size;
}

/* ------------------------------------------------------------- */

static NetClientInfo net_xen_info = {
    .type = NET_CLIENT_TYPE_NIC,
    .size = sizeof(NICState),
    .can_receive = net_rx_ok,
    .receive = net_rx_packet,
};

static int net_init(struct XenDevice *xendev)
{
    struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);

    /* read xenstore entries */
    if (netdev->mac == NULL) {
        netdev->mac = xenstore_read_be_str(&netdev->xendev, "mac");
    }

    /* do we have all we need? */
    if (netdev->mac == NULL) {
        return -1;
    }

    if (net_parse_macaddr(netdev->conf.macaddr.a, netdev->mac) < 0) {
        return -1;
    }

    netdev->conf.vlan = qemu_find_vlan(netdev->xendev.dev, 1);
    netdev->conf.peer = NULL;

    netdev->nic = qemu_new_nic(&net_xen_info, &netdev->conf,
                               "xen", NULL, netdev);

    snprintf(netdev->nic->nc.info_str, sizeof(netdev->nic->nc.info_str),
             "nic: xenbus vif macaddr=%s", netdev->mac);

    /* fill info */
    xenstore_write_be_int(&netdev->xendev, "feature-rx-copy", 1);
    xenstore_write_be_int(&netdev->xendev, "feature-rx-flip", 0);

    return 0;
}

static int net_connect(struct XenDevice *xendev)
{
    struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
    int rx_copy;

    if (xenstore_read_fe_int(&netdev->xendev, "tx-ring-ref",
                             &netdev->tx_ring_ref) == -1) {
        return -1;
    }
    if (xenstore_read_fe_int(&netdev->xendev, "rx-ring-ref",
                             &netdev->rx_ring_ref) == -1) {
        return 1;
    }
    if (xenstore_read_fe_int(&netdev->xendev, "event-channel",
                             &netdev->xendev.remote_port) == -1) {
        return -1;
    }

    if (xenstore_read_fe_int(&netdev->xendev, "request-rx-copy", &rx_copy) == -1) {
        rx_copy = 0;
    }
    if (rx_copy == 0) {
        xen_be_printf(&netdev->xendev, 0, "frontend doesn't support rx-copy.\n");
        return -1;
    }

    netdev->txs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
                                          netdev->xendev.dom,
                                          netdev->tx_ring_ref,
                                          PROT_READ | PROT_WRITE);
    netdev->rxs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
                                          netdev->xendev.dom,
                                          netdev->rx_ring_ref,
                                          PROT_READ | PROT_WRITE);
    if (!netdev->txs || !netdev->rxs) {
        return -1;
    }
    BACK_RING_INIT(&netdev->tx_ring, netdev->txs, XC_PAGE_SIZE);
    BACK_RING_INIT(&netdev->rx_ring, netdev->rxs, XC_PAGE_SIZE);

    xen_be_bind_evtchn(&netdev->xendev);

    xen_be_printf(&netdev->xendev, 1, "ok: tx-ring-ref %d, rx-ring-ref %d, "
                  "remote port %d, local port %d\n",
                  netdev->tx_ring_ref, netdev->rx_ring_ref,
                  netdev->xendev.remote_port, netdev->xendev.local_port);

    net_tx_packets(netdev);
    return 0;
}

static void net_disconnect(struct XenDevice *xendev)
{
    struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);

    xen_be_unbind_evtchn(&netdev->xendev);

    if (netdev->txs) {
        xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->txs, 1);
        netdev->txs = NULL;
    }
    if (netdev->rxs) {
        xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->rxs, 1);
        netdev->rxs = NULL;
    }
    if (netdev->nic) {
        qemu_del_vlan_client(&netdev->nic->nc);
        netdev->nic = NULL;
    }
}

static void net_event(struct XenDevice *xendev)
{
    struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
    net_tx_packets(netdev);
}

static int net_free(struct XenDevice *xendev)
{
    struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);

    qemu_free(netdev->mac);
    return 0;
}

/* ------------------------------------------------------------- */

struct XenDevOps xen_netdev_ops = {
    .size       = sizeof(struct XenNetDev),
    .flags      = DEVOPS_FLAG_NEED_GNTDEV,
    .init       = net_init,
    .connect    = net_connect,
    .event      = net_event,
    .disconnect = net_disconnect,
    .free       = net_free,
};
Commit	Line	Data
e613b064 AL	1	/*
	2	* xen paravirt network card backend
	3	*
	4	* (c) Gerd Hoffmann <kraxel@redhat.com>
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License as published by
	8	* the Free Software Foundation; under version 2 of the License.
	9	*
	10	* This program is distributed in the hope that it will be useful,
	11	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	12	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	13	* GNU General Public License for more details.
	14	*
	15	* You should have received a copy of the GNU General Public License along
8167ee88	16	* with this program; if not, see <http://www.gnu.org/licenses/>.
e613b064 AL	17	*/
	18
	19	#include <stdio.h>
	20	#include <stdlib.h>
	21	#include <stdarg.h>
	22	#include <string.h>
	23	#include <unistd.h>
	24	#include <signal.h>
	25	#include <inttypes.h>
	26	#include <fcntl.h>
	27	#include <errno.h>
	28	#include <pthread.h>
	29	#include <sys/socket.h>
	30	#include <sys/ioctl.h>
	31	#include <sys/types.h>
	32	#include <sys/stat.h>
	33	#include <sys/mman.h>
	34	#include <sys/wait.h>
e613b064 AL	35
	36	#include <xs.h>
	37	#include <xenctrl.h>
	38	#include <xen/io/xenbus.h>
	39	#include <xen/io/netif.h>
	40
	41	#include "hw.h"
	42	#include "net.h"
7200ac3c	43	#include "net/checksum.h"
658788c5	44	#include "net/util.h"
e613b064 AL	45	#include "qemu-char.h"
	46	#include "xen_backend.h"
	47
	48	/* ------------------------------------------------------------- */
	49
	50	struct XenNetDev {
	51	struct XenDevice xendev; /* must be first */
	52	char *mac;
	53	int tx_work;
	54	int tx_ring_ref;
	55	int rx_ring_ref;
	56	struct netif_tx_sring *txs;
	57	struct netif_rx_sring *rxs;
	58	netif_tx_back_ring_t tx_ring;
	59	netif_rx_back_ring_t rx_ring;
658788c5 MM	60	NICConf conf;
658788c5 MM	61	NICState *nic;
e613b064 AL	62	};
	63
	64	/* ------------------------------------------------------------- */
	65
	66	static void net_tx_response(struct XenNetDev netdev, netif_tx_request_t txp, int8_t st)
	67	{
	68	RING_IDX i = netdev->tx_ring.rsp_prod_pvt;
	69	netif_tx_response_t *resp;
	70	int notify;
	71
	72	resp = RING_GET_RESPONSE(&netdev->tx_ring, i);
	73	resp->id = txp->id;
	74	resp->status = st;
	75
	76	#if 0
209cd7ab AP	77	if (txp->flags & NETTXF_extra_info) {
	78	RING_GET_RESPONSE(&netdev->tx_ring, ++i)->status = NETIF_RSP_NULL;
	79	}
e613b064 AL	80	#endif
	81
	82	netdev->tx_ring.rsp_prod_pvt = ++i;
	83	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->tx_ring, notify);
209cd7ab AP	84	if (notify) {
	85	xen_be_send_notify(&netdev->xendev);
	86	}
e613b064 AL	87
e613b064 AL	88	if (i == netdev->tx_ring.req_cons) {
209cd7ab AP	89	int more_to_do;
	90	RING_FINAL_CHECK_FOR_REQUESTS(&netdev->tx_ring, more_to_do);
	91	if (more_to_do) {
	92	netdev->tx_work++;
	93	}
e613b064 AL	94	}
	95	}
	96
	97	static void net_tx_error(struct XenNetDev netdev, netif_tx_request_t txp, RING_IDX end)
	98	{
	99	#if 0
	100	/*
	101	* Hmm, why netback fails everything in the ring?
	102	* Should we do that even when not supporting SG and TSO?
	103	*/
	104	RING_IDX cons = netdev->tx_ring.req_cons;
	105
	106	do {
209cd7ab AP	107	make_tx_response(netif, txp, NETIF_RSP_ERROR);
	108	if (cons >= end) {
	109	break;
	110	}
	111	txp = RING_GET_REQUEST(&netdev->tx_ring, cons++);
e613b064 AL	112	} while (1);
	113	netdev->tx_ring.req_cons = cons;
	114	netif_schedule_work(netif);
	115	netif_put(netif);
	116	#else
	117	net_tx_response(netdev, txp, NETIF_RSP_ERROR);
	118	#endif
	119	}
	120
	121	static void net_tx_packets(struct XenNetDev *netdev)
	122	{
	123	netif_tx_request_t txreq;
	124	RING_IDX rc, rp;
	125	void *page;
	126	void *tmpbuf = NULL;
	127
	128	for (;;) {
209cd7ab AP	129	rc = netdev->tx_ring.req_cons;
	130	rp = netdev->tx_ring.sring->req_prod;
	131	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
e613b064	132
209cd7ab AP	133	while ((rc != rp)) {
	134	if (RING_REQUEST_CONS_OVERFLOW(&netdev->tx_ring, rc)) {
	135	break;
	136	}
	137	memcpy(&txreq, RING_GET_REQUEST(&netdev->tx_ring, rc), sizeof(txreq));
	138	netdev->tx_ring.req_cons = ++rc;
e613b064 AL	139
e613b064 AL	140	#if 1
209cd7ab AP	141	/* should not happen in theory, we don't announce the *
	142	* feature-{sg,gso,whatelse} flags in xenstore (yet?) */
	143	if (txreq.flags & NETTXF_extra_info) {
	144	xen_be_printf(&netdev->xendev, 0, "FIXME: extra info flag\n");
	145	net_tx_error(netdev, &txreq, rc);
	146	continue;
	147	}
	148	if (txreq.flags & NETTXF_more_data) {
	149	xen_be_printf(&netdev->xendev, 0, "FIXME: more data flag\n");
	150	net_tx_error(netdev, &txreq, rc);
	151	continue;
	152	}
e613b064 AL	153	#endif
e613b064 AL	154
209cd7ab AP	155	if (txreq.size < 14) {
	156	xen_be_printf(&netdev->xendev, 0, "bad packet size: %d\n", txreq.size);
	157	net_tx_error(netdev, &txreq, rc);
	158	continue;
	159	}
	160
	161	if ((txreq.offset + txreq.size) > XC_PAGE_SIZE) {
	162	xen_be_printf(&netdev->xendev, 0, "error: page crossing\n");
	163	net_tx_error(netdev, &txreq, rc);
	164	continue;
	165	}
	166
	167	xen_be_printf(&netdev->xendev, 3, "tx packet ref %d, off %d, len %d, flags 0x%x%s%s%s%s\n",
	168	txreq.gref, txreq.offset, txreq.size, txreq.flags,
	169	(txreq.flags & NETTXF_csum_blank) ? " csum_blank" : "",
	170	(txreq.flags & NETTXF_data_validated) ? " data_validated" : "",
	171	(txreq.flags & NETTXF_more_data) ? " more_data" : "",
	172	(txreq.flags & NETTXF_extra_info) ? " extra_info" : "");
	173
	174	page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
	175	netdev->xendev.dom,
	176	txreq.gref, PROT_READ);
	177	if (page == NULL) {
	178	xen_be_printf(&netdev->xendev, 0, "error: tx gref dereference failed (%d)\n",
e613b064	179	txreq.gref);
209cd7ab AP	180	net_tx_error(netdev, &txreq, rc);
	181	continue;
	182	}
	183	if (txreq.flags & NETTXF_csum_blank) {
e613b064	184	/* have read-only mapping -> can't fill checksum in-place */
209cd7ab	185	if (!tmpbuf) {
682aea0e	186	tmpbuf = qemu_malloc(XC_PAGE_SIZE);
209cd7ab	187	}
e613b064	188	memcpy(tmpbuf, page + txreq.offset, txreq.size);
209cd7ab	189	net_checksum_calculate(tmpbuf, txreq.size);
658788c5	190	qemu_send_packet(&netdev->nic->nc, tmpbuf, txreq.size);
e613b064	191	} else {
658788c5	192	qemu_send_packet(&netdev->nic->nc, page + txreq.offset, txreq.size);
e613b064	193	}
209cd7ab AP	194	xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
	195	net_tx_response(netdev, &txreq, NETIF_RSP_OKAY);
	196	}
	197	if (!netdev->tx_work) {
	198	break;
	199	}
	200	netdev->tx_work = 0;
e613b064	201	}
7105b056	202	qemu_free(tmpbuf);
e613b064 AL	203	}
	204
	205	/* ------------------------------------------------------------- */
	206
	207	static void net_rx_response(struct XenNetDev *netdev,
209cd7ab AP	208	netif_rx_request_t *req, int8_t st,
	209	uint16_t offset, uint16_t size,
	210	uint16_t flags)
e613b064 AL	211	{
	212	RING_IDX i = netdev->rx_ring.rsp_prod_pvt;
	213	netif_rx_response_t *resp;
	214	int notify;
	215
	216	resp = RING_GET_RESPONSE(&netdev->rx_ring, i);
	217	resp->offset = offset;
	218	resp->flags = flags;
	219	resp->id = req->id;
	220	resp->status = (int16_t)size;
209cd7ab AP	221	if (st < 0) {
	222	resp->status = (int16_t)st;
	223	}
e613b064 AL	224
e613b064 AL	225	xen_be_printf(&netdev->xendev, 3, "rx response: idx %d, status %d, flags 0x%x\n",
209cd7ab	226	i, resp->status, resp->flags);
e613b064 AL	227
	228	netdev->rx_ring.rsp_prod_pvt = ++i;
	229	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->rx_ring, notify);
209cd7ab AP	230	if (notify) {
	231	xen_be_send_notify(&netdev->xendev);
	232	}
e613b064 AL	233	}
	234
	235	#define NET_IP_ALIGN 2
	236
658788c5	237	static int net_rx_ok(VLANClientState *nc)
e613b064	238	{
658788c5	239	struct XenNetDev *netdev = DO_UPCAST(NICState, nc, nc)->opaque;
e613b064 AL	240	RING_IDX rc, rp;
e613b064 AL	241
209cd7ab AP	242	if (netdev->xendev.be_state != XenbusStateConnected) {
	243	return 0;
	244	}
e613b064 AL	245
	246	rc = netdev->rx_ring.req_cons;
	247	rp = netdev->rx_ring.sring->req_prod;
	248	xen_rmb();
	249
	250	if (rc == rp \|\| RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
209cd7ab AP	251	xen_be_printf(&netdev->xendev, 2, "%s: no rx buffers (%d/%d)\n",
	252	__FUNCTION__, rc, rp);
	253	return 0;
e613b064 AL	254	}
	255	return 1;
	256	}
	257
658788c5	258	static ssize_t net_rx_packet(VLANClientState nc, const uint8_t buf, size_t size)
e613b064	259	{
658788c5	260	struct XenNetDev *netdev = DO_UPCAST(NICState, nc, nc)->opaque;
e613b064 AL	261	netif_rx_request_t rxreq;
	262	RING_IDX rc, rp;
	263	void *page;
	264
209cd7ab AP	265	if (netdev->xendev.be_state != XenbusStateConnected) {
	266	return -1;
	267	}
e613b064 AL	268
	269	rc = netdev->rx_ring.req_cons;
	270	rp = netdev->rx_ring.sring->req_prod;
	271	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
	272
	273	if (rc == rp \|\| RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
209cd7ab AP	274	xen_be_printf(&netdev->xendev, 2, "no buffer, drop packet\n");
209cd7ab AP	275	return -1;
e613b064 AL	276	}
e613b064 AL	277	if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
209cd7ab AP	278	xen_be_printf(&netdev->xendev, 0, "packet too big (%lu > %ld)",
	279	(unsigned long)size, XC_PAGE_SIZE - NET_IP_ALIGN);
	280	return -1;
e613b064 AL	281	}
	282
	283	memcpy(&rxreq, RING_GET_REQUEST(&netdev->rx_ring, rc), sizeof(rxreq));
	284	netdev->rx_ring.req_cons = ++rc;
	285
	286	page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	287	netdev->xendev.dom,
209cd7ab AP	288	rxreq.gref, PROT_WRITE);
e613b064	289	if (page == NULL) {
209cd7ab	290	xen_be_printf(&netdev->xendev, 0, "error: rx gref dereference failed (%d)\n",
e613b064	291	rxreq.gref);
209cd7ab AP	292	net_rx_response(netdev, &rxreq, NETIF_RSP_ERROR, 0, 0, 0);
209cd7ab AP	293	return -1;
e613b064 AL	294	}
	295	memcpy(page + NET_IP_ALIGN, buf, size);
	296	xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
	297	net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);
4f1c942b MM	298
4f1c942b MM	299	return size;
e613b064 AL	300	}
	301
	302	/* ------------------------------------------------------------- */
	303
658788c5 MM	304	static NetClientInfo net_xen_info = {
	305	.type = NET_CLIENT_TYPE_NIC,
	306	.size = sizeof(NICState),
	307	.can_receive = net_rx_ok,
	308	.receive = net_rx_packet,
	309	};
	310
e613b064 AL	311	static int net_init(struct XenDevice *xendev)
	312	{
	313	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
e613b064 AL	314
e613b064 AL	315	/* read xenstore entries */
209cd7ab AP	316	if (netdev->mac == NULL) {
	317	netdev->mac = xenstore_read_be_str(&netdev->xendev, "mac");
	318	}
e613b064 AL	319
e613b064 AL	320	/* do we have all we need? */
209cd7ab AP	321	if (netdev->mac == NULL) {
	322	return -1;
	323	}
e613b064	324
209cd7ab	325	if (net_parse_macaddr(netdev->conf.macaddr.a, netdev->mac) < 0) {
658788c5	326	return -1;
209cd7ab	327	}
658788c5 MM	328
	329	netdev->conf.vlan = qemu_find_vlan(netdev->xendev.dev, 1);
	330	netdev->conf.peer = NULL;
	331
	332	netdev->nic = qemu_new_nic(&net_xen_info, &netdev->conf,
	333	"xen", NULL, netdev);
	334
	335	snprintf(netdev->nic->nc.info_str, sizeof(netdev->nic->nc.info_str),
e613b064 AL	336	"nic: xenbus vif macaddr=%s", netdev->mac);
	337
	338	/* fill info */
	339	xenstore_write_be_int(&netdev->xendev, "feature-rx-copy", 1);
	340	xenstore_write_be_int(&netdev->xendev, "feature-rx-flip", 0);
	341
	342	return 0;
	343	}
	344
	345	static int net_connect(struct XenDevice *xendev)
	346	{
	347	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	348	int rx_copy;
	349
	350	if (xenstore_read_fe_int(&netdev->xendev, "tx-ring-ref",
209cd7ab AP	351	&netdev->tx_ring_ref) == -1) {
	352	return -1;
	353	}
e613b064	354	if (xenstore_read_fe_int(&netdev->xendev, "rx-ring-ref",
209cd7ab AP	355	&netdev->rx_ring_ref) == -1) {
	356	return 1;
	357	}
e613b064	358	if (xenstore_read_fe_int(&netdev->xendev, "event-channel",
209cd7ab AP	359	&netdev->xendev.remote_port) == -1) {
	360	return -1;
	361	}
e613b064	362
209cd7ab AP	363	if (xenstore_read_fe_int(&netdev->xendev, "request-rx-copy", &rx_copy) == -1) {
	364	rx_copy = 0;
	365	}
e613b064	366	if (rx_copy == 0) {
209cd7ab AP	367	xen_be_printf(&netdev->xendev, 0, "frontend doesn't support rx-copy.\n");
209cd7ab AP	368	return -1;
e613b064 AL	369	}
	370
	371	netdev->txs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	372	netdev->xendev.dom,
	373	netdev->tx_ring_ref,
	374	PROT_READ \| PROT_WRITE);
e613b064	375	netdev->rxs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	376	netdev->xendev.dom,
	377	netdev->rx_ring_ref,
	378	PROT_READ \| PROT_WRITE);
	379	if (!netdev->txs \|\| !netdev->rxs) {
	380	return -1;
	381	}
e613b064 AL	382	BACK_RING_INIT(&netdev->tx_ring, netdev->txs, XC_PAGE_SIZE);
	383	BACK_RING_INIT(&netdev->rx_ring, netdev->rxs, XC_PAGE_SIZE);
	384
	385	xen_be_bind_evtchn(&netdev->xendev);
	386
	387	xen_be_printf(&netdev->xendev, 1, "ok: tx-ring-ref %d, rx-ring-ref %d, "
209cd7ab AP	388	"remote port %d, local port %d\n",
	389	netdev->tx_ring_ref, netdev->rx_ring_ref,
	390	netdev->xendev.remote_port, netdev->xendev.local_port);
3e3cabcf GH	391
3e3cabcf GH	392	net_tx_packets(netdev);
e613b064 AL	393	return 0;
	394	}
	395
	396	static void net_disconnect(struct XenDevice *xendev)
	397	{
	398	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	399
	400	xen_be_unbind_evtchn(&netdev->xendev);
	401
	402	if (netdev->txs) {
209cd7ab AP	403	xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->txs, 1);
209cd7ab AP	404	netdev->txs = NULL;
e613b064 AL	405	}
e613b064 AL	406	if (netdev->rxs) {
209cd7ab AP	407	xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->rxs, 1);
209cd7ab AP	408	netdev->rxs = NULL;
e613b064	409	}
658788c5 MM	410	if (netdev->nic) {
	411	qemu_del_vlan_client(&netdev->nic->nc);
	412	netdev->nic = NULL;
e613b064 AL	413	}
	414	}
	415
	416	static void net_event(struct XenDevice *xendev)
	417	{
	418	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	419	net_tx_packets(netdev);
	420	}
	421
	422	static int net_free(struct XenDevice *xendev)
	423	{
	424	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	425
	426	qemu_free(netdev->mac);
	427	return 0;
	428	}
	429
	430	/* ------------------------------------------------------------- */
	431
	432	struct XenDevOps xen_netdev_ops = {
	433	.size = sizeof(struct XenNetDev),
	434	.flags = DEVOPS_FLAG_NEED_GNTDEV,
	435	.init = net_init,
	436	.connect = net_connect,
	437	.event = net_event,
	438	.disconnect = net_disconnect,
	439	.free = net_free,
	440	};