[mirror_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/>.
 *
 *  Contributions after 2012-01-13 are licensed under the terms of the
 *  GNU GPL, version 2 or (at your option) any later version.
 */

#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 <sys/socket.h>
#include <sys/ioctl.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mman.h>
#include <sys/wait.h>

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

#include <xen/io/netif.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 = g_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;
    }
    g_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_OPTIONS_KIND_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.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);

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

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

struct XenDevOps xen_netdev_ops = {
    .size       = sizeof(struct XenNetDev),
    .flags      = DEVOPS_FLAG_NEED_GNTDEV,
    .init       = net_init,
    .initialise    = 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/>.
6b620ca3 PB	17	*
	18	* Contributions after 2012-01-13 are licensed under the terms of the
	19	* GNU GPL, version 2 or (at your option) any later version.
e613b064 AL	20	*/
	21
	22	#include <stdio.h>
	23	#include <stdlib.h>
	24	#include <stdarg.h>
	25	#include <string.h>
	26	#include <unistd.h>
	27	#include <signal.h>
	28	#include <inttypes.h>
	29	#include <fcntl.h>
	30	#include <errno.h>
e613b064 AL	31	#include <sys/socket.h>
	32	#include <sys/ioctl.h>
	33	#include <sys/types.h>
	34	#include <sys/stat.h>
	35	#include <sys/mman.h>
	36	#include <sys/wait.h>
e613b064	37
e613b064 AL	38	#include "hw.h"
e613b064 AL	39	#include "net.h"
7200ac3c	40	#include "net/checksum.h"
658788c5	41	#include "net/util.h"
e613b064 AL	42	#include "qemu-char.h"
	43	#include "xen_backend.h"
	44
b41f6719 AP	45	#include <xen/io/netif.h>
b41f6719 AP	46
e613b064 AL	47	/* ------------------------------------------------------------- */
	48
	49	struct XenNetDev {
	50	struct XenDevice xendev; /* must be first */
	51	char *mac;
	52	int tx_work;
	53	int tx_ring_ref;
	54	int rx_ring_ref;
	55	struct netif_tx_sring *txs;
	56	struct netif_rx_sring *rxs;
	57	netif_tx_back_ring_t tx_ring;
	58	netif_rx_back_ring_t rx_ring;
658788c5 MM	59	NICConf conf;
658788c5 MM	60	NICState *nic;
e613b064 AL	61	};
	62
	63	/* ------------------------------------------------------------- */
	64
	65	static void net_tx_response(struct XenNetDev netdev, netif_tx_request_t txp, int8_t st)
	66	{
	67	RING_IDX i = netdev->tx_ring.rsp_prod_pvt;
	68	netif_tx_response_t *resp;
	69	int notify;
	70
	71	resp = RING_GET_RESPONSE(&netdev->tx_ring, i);
	72	resp->id = txp->id;
	73	resp->status = st;
	74
	75	#if 0
209cd7ab AP	76	if (txp->flags & NETTXF_extra_info) {
	77	RING_GET_RESPONSE(&netdev->tx_ring, ++i)->status = NETIF_RSP_NULL;
	78	}
e613b064 AL	79	#endif
	80
	81	netdev->tx_ring.rsp_prod_pvt = ++i;
	82	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->tx_ring, notify);
209cd7ab AP	83	if (notify) {
	84	xen_be_send_notify(&netdev->xendev);
	85	}
e613b064 AL	86
e613b064 AL	87	if (i == netdev->tx_ring.req_cons) {
209cd7ab AP	88	int more_to_do;
	89	RING_FINAL_CHECK_FOR_REQUESTS(&netdev->tx_ring, more_to_do);
	90	if (more_to_do) {
	91	netdev->tx_work++;
	92	}
e613b064 AL	93	}
	94	}
	95
	96	static void net_tx_error(struct XenNetDev netdev, netif_tx_request_t txp, RING_IDX end)
	97	{
	98	#if 0
	99	/*
	100	* Hmm, why netback fails everything in the ring?
	101	* Should we do that even when not supporting SG and TSO?
	102	*/
	103	RING_IDX cons = netdev->tx_ring.req_cons;
	104
	105	do {
209cd7ab AP	106	make_tx_response(netif, txp, NETIF_RSP_ERROR);
	107	if (cons >= end) {
	108	break;
	109	}
	110	txp = RING_GET_REQUEST(&netdev->tx_ring, cons++);
e613b064 AL	111	} while (1);
	112	netdev->tx_ring.req_cons = cons;
	113	netif_schedule_work(netif);
	114	netif_put(netif);
	115	#else
	116	net_tx_response(netdev, txp, NETIF_RSP_ERROR);
	117	#endif
	118	}
	119
	120	static void net_tx_packets(struct XenNetDev *netdev)
	121	{
	122	netif_tx_request_t txreq;
	123	RING_IDX rc, rp;
	124	void *page;
	125	void *tmpbuf = NULL;
	126
	127	for (;;) {
209cd7ab AP	128	rc = netdev->tx_ring.req_cons;
	129	rp = netdev->tx_ring.sring->req_prod;
	130	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
e613b064	131
209cd7ab AP	132	while ((rc != rp)) {
	133	if (RING_REQUEST_CONS_OVERFLOW(&netdev->tx_ring, rc)) {
	134	break;
	135	}
	136	memcpy(&txreq, RING_GET_REQUEST(&netdev->tx_ring, rc), sizeof(txreq));
	137	netdev->tx_ring.req_cons = ++rc;
e613b064 AL	138
e613b064 AL	139	#if 1
209cd7ab AP	140	/* should not happen in theory, we don't announce the *
	141	* feature-{sg,gso,whatelse} flags in xenstore (yet?) */
	142	if (txreq.flags & NETTXF_extra_info) {
	143	xen_be_printf(&netdev->xendev, 0, "FIXME: extra info flag\n");
	144	net_tx_error(netdev, &txreq, rc);
	145	continue;
	146	}
	147	if (txreq.flags & NETTXF_more_data) {
	148	xen_be_printf(&netdev->xendev, 0, "FIXME: more data flag\n");
	149	net_tx_error(netdev, &txreq, rc);
	150	continue;
	151	}
e613b064 AL	152	#endif
e613b064 AL	153
209cd7ab AP	154	if (txreq.size < 14) {
	155	xen_be_printf(&netdev->xendev, 0, "bad packet size: %d\n", txreq.size);
	156	net_tx_error(netdev, &txreq, rc);
	157	continue;
	158	}
	159
	160	if ((txreq.offset + txreq.size) > XC_PAGE_SIZE) {
	161	xen_be_printf(&netdev->xendev, 0, "error: page crossing\n");
	162	net_tx_error(netdev, &txreq, rc);
	163	continue;
	164	}
	165
	166	xen_be_printf(&netdev->xendev, 3, "tx packet ref %d, off %d, len %d, flags 0x%x%s%s%s%s\n",
	167	txreq.gref, txreq.offset, txreq.size, txreq.flags,
	168	(txreq.flags & NETTXF_csum_blank) ? " csum_blank" : "",
	169	(txreq.flags & NETTXF_data_validated) ? " data_validated" : "",
	170	(txreq.flags & NETTXF_more_data) ? " more_data" : "",
	171	(txreq.flags & NETTXF_extra_info) ? " extra_info" : "");
	172
	173	page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
	174	netdev->xendev.dom,
	175	txreq.gref, PROT_READ);
	176	if (page == NULL) {
	177	xen_be_printf(&netdev->xendev, 0, "error: tx gref dereference failed (%d)\n",
e613b064	178	txreq.gref);
209cd7ab AP	179	net_tx_error(netdev, &txreq, rc);
	180	continue;
	181	}
	182	if (txreq.flags & NETTXF_csum_blank) {
e613b064	183	/* have read-only mapping -> can't fill checksum in-place */
209cd7ab	184	if (!tmpbuf) {
7267c094	185	tmpbuf = g_malloc(XC_PAGE_SIZE);
209cd7ab	186	}
e613b064	187	memcpy(tmpbuf, page + txreq.offset, txreq.size);
209cd7ab	188	net_checksum_calculate(tmpbuf, txreq.size);
658788c5	189	qemu_send_packet(&netdev->nic->nc, tmpbuf, txreq.size);
e613b064	190	} else {
658788c5	191	qemu_send_packet(&netdev->nic->nc, page + txreq.offset, txreq.size);
e613b064	192	}
209cd7ab AP	193	xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
	194	net_tx_response(netdev, &txreq, NETIF_RSP_OKAY);
	195	}
	196	if (!netdev->tx_work) {
	197	break;
	198	}
	199	netdev->tx_work = 0;
e613b064	200	}
7267c094	201	g_free(tmpbuf);
e613b064 AL	202	}
	203
	204	/* ------------------------------------------------------------- */
	205
	206	static void net_rx_response(struct XenNetDev *netdev,
209cd7ab AP	207	netif_rx_request_t *req, int8_t st,
	208	uint16_t offset, uint16_t size,
	209	uint16_t flags)
e613b064 AL	210	{
	211	RING_IDX i = netdev->rx_ring.rsp_prod_pvt;
	212	netif_rx_response_t *resp;
	213	int notify;
	214
	215	resp = RING_GET_RESPONSE(&netdev->rx_ring, i);
	216	resp->offset = offset;
	217	resp->flags = flags;
	218	resp->id = req->id;
	219	resp->status = (int16_t)size;
209cd7ab AP	220	if (st < 0) {
	221	resp->status = (int16_t)st;
	222	}
e613b064 AL	223
e613b064 AL	224	xen_be_printf(&netdev->xendev, 3, "rx response: idx %d, status %d, flags 0x%x\n",
209cd7ab	225	i, resp->status, resp->flags);
e613b064 AL	226
	227	netdev->rx_ring.rsp_prod_pvt = ++i;
	228	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->rx_ring, notify);
209cd7ab AP	229	if (notify) {
	230	xen_be_send_notify(&netdev->xendev);
	231	}
e613b064 AL	232	}
	233
	234	#define NET_IP_ALIGN 2
	235
658788c5	236	static int net_rx_ok(VLANClientState *nc)
e613b064	237	{
658788c5	238	struct XenNetDev *netdev = DO_UPCAST(NICState, nc, nc)->opaque;
e613b064 AL	239	RING_IDX rc, rp;
e613b064 AL	240
209cd7ab AP	241	if (netdev->xendev.be_state != XenbusStateConnected) {
	242	return 0;
	243	}
e613b064 AL	244
	245	rc = netdev->rx_ring.req_cons;
	246	rp = netdev->rx_ring.sring->req_prod;
	247	xen_rmb();
	248
	249	if (rc == rp \|\| RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
209cd7ab AP	250	xen_be_printf(&netdev->xendev, 2, "%s: no rx buffers (%d/%d)\n",
	251	__FUNCTION__, rc, rp);
	252	return 0;
e613b064 AL	253	}
	254	return 1;
	255	}
	256
658788c5	257	static ssize_t net_rx_packet(VLANClientState nc, const uint8_t buf, size_t size)
e613b064	258	{
658788c5	259	struct XenNetDev *netdev = DO_UPCAST(NICState, nc, nc)->opaque;
e613b064 AL	260	netif_rx_request_t rxreq;
	261	RING_IDX rc, rp;
	262	void *page;
	263
209cd7ab AP	264	if (netdev->xendev.be_state != XenbusStateConnected) {
	265	return -1;
	266	}
e613b064 AL	267
	268	rc = netdev->rx_ring.req_cons;
	269	rp = netdev->rx_ring.sring->req_prod;
	270	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
	271
	272	if (rc == rp \|\| RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
209cd7ab AP	273	xen_be_printf(&netdev->xendev, 2, "no buffer, drop packet\n");
209cd7ab AP	274	return -1;
e613b064 AL	275	}
e613b064 AL	276	if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
209cd7ab AP	277	xen_be_printf(&netdev->xendev, 0, "packet too big (%lu > %ld)",
	278	(unsigned long)size, XC_PAGE_SIZE - NET_IP_ALIGN);
	279	return -1;
e613b064 AL	280	}
	281
	282	memcpy(&rxreq, RING_GET_REQUEST(&netdev->rx_ring, rc), sizeof(rxreq));
	283	netdev->rx_ring.req_cons = ++rc;
	284
	285	page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	286	netdev->xendev.dom,
209cd7ab AP	287	rxreq.gref, PROT_WRITE);
e613b064	288	if (page == NULL) {
209cd7ab	289	xen_be_printf(&netdev->xendev, 0, "error: rx gref dereference failed (%d)\n",
e613b064	290	rxreq.gref);
209cd7ab AP	291	net_rx_response(netdev, &rxreq, NETIF_RSP_ERROR, 0, 0, 0);
209cd7ab AP	292	return -1;
e613b064 AL	293	}
	294	memcpy(page + NET_IP_ALIGN, buf, size);
	295	xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
	296	net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);
4f1c942b MM	297
4f1c942b MM	298	return size;
e613b064 AL	299	}
	300
	301	/* ------------------------------------------------------------- */
	302
658788c5	303	static NetClientInfo net_xen_info = {
2be64a68	304	.type = NET_CLIENT_OPTIONS_KIND_NIC,
658788c5 MM	305	.size = sizeof(NICState),
	306	.can_receive = net_rx_ok,
	307	.receive = net_rx_packet,
	308	};
	309
e613b064 AL	310	static int net_init(struct XenDevice *xendev)
	311	{
	312	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
e613b064 AL	313
e613b064 AL	314	/* read xenstore entries */
209cd7ab AP	315	if (netdev->mac == NULL) {
	316	netdev->mac = xenstore_read_be_str(&netdev->xendev, "mac");
	317	}
e613b064 AL	318
e613b064 AL	319	/* do we have all we need? */
209cd7ab AP	320	if (netdev->mac == NULL) {
	321	return -1;
	322	}
e613b064	323
209cd7ab	324	if (net_parse_macaddr(netdev->conf.macaddr.a, netdev->mac) < 0) {
658788c5	325	return -1;
209cd7ab	326	}
658788c5	327
658788c5 MM	328	netdev->conf.peer = NULL;
	329
	330	netdev->nic = qemu_new_nic(&net_xen_info, &netdev->conf,
	331	"xen", NULL, netdev);
	332
	333	snprintf(netdev->nic->nc.info_str, sizeof(netdev->nic->nc.info_str),
e613b064 AL	334	"nic: xenbus vif macaddr=%s", netdev->mac);
	335
	336	/* fill info */
	337	xenstore_write_be_int(&netdev->xendev, "feature-rx-copy", 1);
	338	xenstore_write_be_int(&netdev->xendev, "feature-rx-flip", 0);
	339
	340	return 0;
	341	}
	342
	343	static int net_connect(struct XenDevice *xendev)
	344	{
	345	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	346	int rx_copy;
	347
	348	if (xenstore_read_fe_int(&netdev->xendev, "tx-ring-ref",
209cd7ab AP	349	&netdev->tx_ring_ref) == -1) {
	350	return -1;
	351	}
e613b064	352	if (xenstore_read_fe_int(&netdev->xendev, "rx-ring-ref",
209cd7ab AP	353	&netdev->rx_ring_ref) == -1) {
	354	return 1;
	355	}
e613b064	356	if (xenstore_read_fe_int(&netdev->xendev, "event-channel",
209cd7ab AP	357	&netdev->xendev.remote_port) == -1) {
	358	return -1;
	359	}
e613b064	360
209cd7ab AP	361	if (xenstore_read_fe_int(&netdev->xendev, "request-rx-copy", &rx_copy) == -1) {
	362	rx_copy = 0;
	363	}
e613b064	364	if (rx_copy == 0) {
209cd7ab AP	365	xen_be_printf(&netdev->xendev, 0, "frontend doesn't support rx-copy.\n");
209cd7ab AP	366	return -1;
e613b064 AL	367	}
	368
	369	netdev->txs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	370	netdev->xendev.dom,
	371	netdev->tx_ring_ref,
	372	PROT_READ \| PROT_WRITE);
e613b064	373	netdev->rxs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	374	netdev->xendev.dom,
	375	netdev->rx_ring_ref,
	376	PROT_READ \| PROT_WRITE);
	377	if (!netdev->txs \|\| !netdev->rxs) {
	378	return -1;
	379	}
e613b064 AL	380	BACK_RING_INIT(&netdev->tx_ring, netdev->txs, XC_PAGE_SIZE);
	381	BACK_RING_INIT(&netdev->rx_ring, netdev->rxs, XC_PAGE_SIZE);
	382
	383	xen_be_bind_evtchn(&netdev->xendev);
	384
	385	xen_be_printf(&netdev->xendev, 1, "ok: tx-ring-ref %d, rx-ring-ref %d, "
209cd7ab AP	386	"remote port %d, local port %d\n",
	387	netdev->tx_ring_ref, netdev->rx_ring_ref,
	388	netdev->xendev.remote_port, netdev->xendev.local_port);
3e3cabcf GH	389
3e3cabcf GH	390	net_tx_packets(netdev);
e613b064 AL	391	return 0;
	392	}
	393
	394	static void net_disconnect(struct XenDevice *xendev)
	395	{
	396	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	397
	398	xen_be_unbind_evtchn(&netdev->xendev);
	399
	400	if (netdev->txs) {
209cd7ab AP	401	xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->txs, 1);
209cd7ab AP	402	netdev->txs = NULL;
e613b064 AL	403	}
e613b064 AL	404	if (netdev->rxs) {
209cd7ab AP	405	xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->rxs, 1);
209cd7ab AP	406	netdev->rxs = NULL;
e613b064	407	}
658788c5 MM	408	if (netdev->nic) {
	409	qemu_del_vlan_client(&netdev->nic->nc);
	410	netdev->nic = NULL;
e613b064 AL	411	}
	412	}
	413
	414	static void net_event(struct XenDevice *xendev)
	415	{
	416	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	417	net_tx_packets(netdev);
	418	}
	419
	420	static int net_free(struct XenDevice *xendev)
	421	{
	422	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	423
7267c094	424	g_free(netdev->mac);
e613b064 AL	425	return 0;
	426	}
	427
	428	/* ------------------------------------------------------------- */
	429
	430	struct XenDevOps xen_netdev_ops = {
	431	.size = sizeof(struct XenNetDev),
	432	.flags = DEVOPS_FLAG_NEED_GNTDEV,
	433	.init = net_init,
384087b2	434	.initialise = net_connect,
e613b064 AL	435	.event = net_event,
	436	.disconnect = net_disconnect,
	437	.free = net_free,
	438	};