[thirdparty/qemu.git] / hw / net / 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 "qemu/osdep.h"
#include <sys/socket.h>
#include <sys/ioctl.h>
#include <sys/wait.h>

#include "net/net.h"
#include "net/checksum.h"
#include "net/util.h"
#include "hw/xen/xen-legacy-backend.h"

#include "hw/xen/interface/io/netif.h"

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

struct XenNetDev {
    struct XenLegacyDevice      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_pv_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_pv_printf(&netdev->xendev, 0, "FIXME: extra info flag\n");
                net_tx_error(netdev, &txreq, rc);
                continue;
            }
            if (txreq.flags & NETTXF_more_data) {
                xen_pv_printf(&netdev->xendev, 0, "FIXME: more data flag\n");
                net_tx_error(netdev, &txreq, rc);
                continue;
            }
#endif

            if (txreq.size < 14) {
                xen_pv_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_pv_printf(&netdev->xendev, 0, "error: page crossing\n");
                net_tx_error(netdev, &txreq, rc);
                continue;
            }

            xen_pv_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 = xen_be_map_grant_ref(&netdev->xendev, txreq.gref,
                                        PROT_READ);
            if (page == NULL) {
                xen_pv_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(qemu_get_queue(netdev->nic), tmpbuf,
                                 txreq.size);
            } else {
                qemu_send_packet(qemu_get_queue(netdev->nic),
                                 page + txreq.offset, txreq.size);
            }
            xen_be_unmap_grant_ref(&netdev->xendev, page);
            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_pv_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_pv_send_notify(&netdev->xendev);
    }
}

#define NET_IP_ALIGN 2

static ssize_t net_rx_packet(NetClientState *nc, const uint8_t *buf, size_t size)
{
    struct XenNetDev *netdev = qemu_get_nic_opaque(nc);
    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)) {
        return 0;
    }
    if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
        xen_pv_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 = xen_be_map_grant_ref(&netdev->xendev, rxreq.gref, PROT_WRITE);
    if (page == NULL) {
        xen_pv_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);
    xen_be_unmap_grant_ref(&netdev->xendev, page);
    net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);

    return size;
}

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

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

static int net_init(struct XenLegacyDevice *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->nic = qemu_new_nic(&net_xen_info, &netdev->conf,
                               "xen", NULL, netdev);

    snprintf(qemu_get_queue(netdev->nic)->info_str,
             sizeof(qemu_get_queue(netdev->nic)->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 XenLegacyDevice *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_pv_printf(&netdev->xendev, 0,
                      "frontend doesn't support rx-copy.\n");
        return -1;
    }

    netdev->txs = xen_be_map_grant_ref(&netdev->xendev,
                                       netdev->tx_ring_ref,
                                       PROT_READ | PROT_WRITE);
    if (!netdev->txs) {
        return -1;
    }
    netdev->rxs = xen_be_map_grant_ref(&netdev->xendev,
                                       netdev->rx_ring_ref,
                                       PROT_READ | PROT_WRITE);
    if (!netdev->rxs) {
        xen_be_unmap_grant_ref(&netdev->xendev, netdev->txs);
        netdev->txs = NULL;
        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_pv_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 XenLegacyDevice *xendev)
{
    struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);

    xen_pv_unbind_evtchn(&netdev->xendev);

    if (netdev->txs) {
        xen_be_unmap_grant_ref(&netdev->xendev, netdev->txs);
        netdev->txs = NULL;
    }
    if (netdev->rxs) {
        xen_be_unmap_grant_ref(&netdev->xendev, netdev->rxs);
        netdev->rxs = NULL;
    }
}

static void net_event(struct XenLegacyDevice *xendev)
{
    struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
    net_tx_packets(netdev);
    qemu_flush_queued_packets(qemu_get_queue(netdev->nic));
}

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

    if (netdev->nic) {
        qemu_del_nic(netdev->nic);
        netdev->nic = NULL;
    }
    g_free(netdev->mac);
    netdev->mac = NULL;
    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	*/
e613b064 AL	21
21cbfe5f	22	#include "qemu/osdep.h"
e613b064 AL	23	#include <sys/socket.h>
e613b064 AL	24	#include <sys/ioctl.h>
e613b064	25	#include <sys/wait.h>
e613b064	26
1422e32d	27	#include "net/net.h"
7200ac3c	28	#include "net/checksum.h"
658788c5	29	#include "net/util.h"
2d0ed5e6	30	#include "hw/xen/xen-legacy-backend.h"
e613b064	31
a3434a2d	32	#include "hw/xen/interface/io/netif.h"
b41f6719	33
e613b064 AL	34	/* ------------------------------------------------------------- */
	35
	36	struct XenNetDev {
2d0ed5e6	37	struct XenLegacyDevice xendev; /* must be first */
e613b064 AL	38	char *mac;
	39	int tx_work;
	40	int tx_ring_ref;
	41	int rx_ring_ref;
	42	struct netif_tx_sring *txs;
	43	struct netif_rx_sring *rxs;
	44	netif_tx_back_ring_t tx_ring;
	45	netif_rx_back_ring_t rx_ring;
658788c5 MM	46	NICConf conf;
658788c5 MM	47	NICState *nic;
e613b064 AL	48	};
	49
	50	/* ------------------------------------------------------------- */
	51
	52	static void net_tx_response(struct XenNetDev netdev, netif_tx_request_t txp, int8_t st)
	53	{
	54	RING_IDX i = netdev->tx_ring.rsp_prod_pvt;
	55	netif_tx_response_t *resp;
	56	int notify;
	57
	58	resp = RING_GET_RESPONSE(&netdev->tx_ring, i);
	59	resp->id = txp->id;
	60	resp->status = st;
	61
	62	#if 0
209cd7ab AP	63	if (txp->flags & NETTXF_extra_info) {
	64	RING_GET_RESPONSE(&netdev->tx_ring, ++i)->status = NETIF_RSP_NULL;
	65	}
e613b064 AL	66	#endif
	67
	68	netdev->tx_ring.rsp_prod_pvt = ++i;
	69	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->tx_ring, notify);
209cd7ab	70	if (notify) {
ba18fa2a	71	xen_pv_send_notify(&netdev->xendev);
209cd7ab	72	}
e613b064 AL	73
e613b064 AL	74	if (i == netdev->tx_ring.req_cons) {
209cd7ab AP	75	int more_to_do;
	76	RING_FINAL_CHECK_FOR_REQUESTS(&netdev->tx_ring, more_to_do);
	77	if (more_to_do) {
	78	netdev->tx_work++;
	79	}
e613b064 AL	80	}
	81	}
	82
	83	static void net_tx_error(struct XenNetDev netdev, netif_tx_request_t txp, RING_IDX end)
	84	{
	85	#if 0
	86	/*
	87	* Hmm, why netback fails everything in the ring?
	88	* Should we do that even when not supporting SG and TSO?
	89	*/
	90	RING_IDX cons = netdev->tx_ring.req_cons;
	91
	92	do {
209cd7ab AP	93	make_tx_response(netif, txp, NETIF_RSP_ERROR);
	94	if (cons >= end) {
	95	break;
	96	}
	97	txp = RING_GET_REQUEST(&netdev->tx_ring, cons++);
e613b064 AL	98	} while (1);
	99	netdev->tx_ring.req_cons = cons;
	100	netif_schedule_work(netif);
	101	netif_put(netif);
	102	#else
	103	net_tx_response(netdev, txp, NETIF_RSP_ERROR);
	104	#endif
	105	}
	106
	107	static void net_tx_packets(struct XenNetDev *netdev)
	108	{
	109	netif_tx_request_t txreq;
	110	RING_IDX rc, rp;
	111	void *page;
	112	void *tmpbuf = NULL;
	113
	114	for (;;) {
209cd7ab AP	115	rc = netdev->tx_ring.req_cons;
	116	rp = netdev->tx_ring.sring->req_prod;
	117	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
e613b064	118
209cd7ab AP	119	while ((rc != rp)) {
	120	if (RING_REQUEST_CONS_OVERFLOW(&netdev->tx_ring, rc)) {
	121	break;
	122	}
	123	memcpy(&txreq, RING_GET_REQUEST(&netdev->tx_ring, rc), sizeof(txreq));
	124	netdev->tx_ring.req_cons = ++rc;
e613b064 AL	125
e613b064 AL	126	#if 1
209cd7ab AP	127	/* should not happen in theory, we don't announce the *
	128	* feature-{sg,gso,whatelse} flags in xenstore (yet?) */
	129	if (txreq.flags & NETTXF_extra_info) {
96c77dba	130	xen_pv_printf(&netdev->xendev, 0, "FIXME: extra info flag\n");
209cd7ab AP	131	net_tx_error(netdev, &txreq, rc);
	132	continue;
	133	}
	134	if (txreq.flags & NETTXF_more_data) {
96c77dba	135	xen_pv_printf(&netdev->xendev, 0, "FIXME: more data flag\n");
209cd7ab AP	136	net_tx_error(netdev, &txreq, rc);
	137	continue;
	138	}
e613b064 AL	139	#endif
e613b064 AL	140
209cd7ab	141	if (txreq.size < 14) {
96c77dba	142	xen_pv_printf(&netdev->xendev, 0, "bad packet size: %d\n",
b9730c5b	143	txreq.size);
209cd7ab AP	144	net_tx_error(netdev, &txreq, rc);
	145	continue;
	146	}
	147
	148	if ((txreq.offset + txreq.size) > XC_PAGE_SIZE) {
96c77dba	149	xen_pv_printf(&netdev->xendev, 0, "error: page crossing\n");
209cd7ab AP	150	net_tx_error(netdev, &txreq, rc);
	151	continue;
	152	}
	153
96c77dba	154	xen_pv_printf(&netdev->xendev, 3,
c22e91b1	155	"tx packet ref %d, off %d, len %d, flags 0x%x%s%s%s%s\n",
209cd7ab AP	156	txreq.gref, txreq.offset, txreq.size, txreq.flags,
	157	(txreq.flags & NETTXF_csum_blank) ? " csum_blank" : "",
	158	(txreq.flags & NETTXF_data_validated) ? " data_validated" : "",
	159	(txreq.flags & NETTXF_more_data) ? " more_data" : "",
	160	(txreq.flags & NETTXF_extra_info) ? " extra_info" : "");
	161
58560f2a PD	162	page = xen_be_map_grant_ref(&netdev->xendev, txreq.gref,
58560f2a PD	163	PROT_READ);
209cd7ab	164	if (page == NULL) {
96c77dba	165	xen_pv_printf(&netdev->xendev, 0,
c22e91b1 EC	166	"error: tx gref dereference failed (%d)\n",
c22e91b1 EC	167	txreq.gref);
209cd7ab AP	168	net_tx_error(netdev, &txreq, rc);
	169	continue;
	170	}
	171	if (txreq.flags & NETTXF_csum_blank) {
e613b064	172	/* have read-only mapping -> can't fill checksum in-place */
209cd7ab	173	if (!tmpbuf) {
7267c094	174	tmpbuf = g_malloc(XC_PAGE_SIZE);
209cd7ab	175	}
e613b064	176	memcpy(tmpbuf, page + txreq.offset, txreq.size);
209cd7ab	177	net_checksum_calculate(tmpbuf, txreq.size);
b356f76d JW	178	qemu_send_packet(qemu_get_queue(netdev->nic), tmpbuf,
b356f76d JW	179	txreq.size);
e613b064	180	} else {
b356f76d JW	181	qemu_send_packet(qemu_get_queue(netdev->nic),
b356f76d JW	182	page + txreq.offset, txreq.size);
e613b064	183	}
58560f2a	184	xen_be_unmap_grant_ref(&netdev->xendev, page);
209cd7ab AP	185	net_tx_response(netdev, &txreq, NETIF_RSP_OKAY);
	186	}
	187	if (!netdev->tx_work) {
	188	break;
	189	}
	190	netdev->tx_work = 0;
e613b064	191	}
7267c094	192	g_free(tmpbuf);
e613b064 AL	193	}
	194
	195	/* ------------------------------------------------------------- */
	196
	197	static void net_rx_response(struct XenNetDev *netdev,
209cd7ab AP	198	netif_rx_request_t *req, int8_t st,
	199	uint16_t offset, uint16_t size,
	200	uint16_t flags)
e613b064 AL	201	{
	202	RING_IDX i = netdev->rx_ring.rsp_prod_pvt;
	203	netif_rx_response_t *resp;
	204	int notify;
	205
	206	resp = RING_GET_RESPONSE(&netdev->rx_ring, i);
	207	resp->offset = offset;
	208	resp->flags = flags;
	209	resp->id = req->id;
	210	resp->status = (int16_t)size;
209cd7ab AP	211	if (st < 0) {
	212	resp->status = (int16_t)st;
	213	}
e613b064	214
96c77dba	215	xen_pv_printf(&netdev->xendev, 3,
b9730c5b	216	"rx response: idx %d, status %d, flags 0x%x\n",
209cd7ab	217	i, resp->status, resp->flags);
e613b064 AL	218
	219	netdev->rx_ring.rsp_prod_pvt = ++i;
	220	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->rx_ring, notify);
209cd7ab	221	if (notify) {
ba18fa2a	222	xen_pv_send_notify(&netdev->xendev);
209cd7ab	223	}
e613b064 AL	224	}
	225
	226	#define NET_IP_ALIGN 2
	227
4e68f7a0	228	static ssize_t net_rx_packet(NetClientState nc, const uint8_t buf, size_t size)
e613b064	229	{
cc1f0f45	230	struct XenNetDev *netdev = qemu_get_nic_opaque(nc);
e613b064 AL	231	netif_rx_request_t rxreq;
	232	RING_IDX rc, rp;
	233	void *page;
	234
209cd7ab AP	235	if (netdev->xendev.be_state != XenbusStateConnected) {
	236	return -1;
	237	}
e613b064 AL	238
	239	rc = netdev->rx_ring.req_cons;
	240	rp = netdev->rx_ring.sring->req_prod;
	241	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
	242
	243	if (rc == rp \|\| RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
7bba83bf	244	return 0;
e613b064 AL	245	}
e613b064 AL	246	if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
96c77dba	247	xen_pv_printf(&netdev->xendev, 0, "packet too big (%lu > %ld)",
209cd7ab AP	248	(unsigned long)size, XC_PAGE_SIZE - NET_IP_ALIGN);
209cd7ab AP	249	return -1;
e613b064 AL	250	}
	251
	252	memcpy(&rxreq, RING_GET_REQUEST(&netdev->rx_ring, rc), sizeof(rxreq));
	253	netdev->rx_ring.req_cons = ++rc;
	254
58560f2a	255	page = xen_be_map_grant_ref(&netdev->xendev, rxreq.gref, PROT_WRITE);
e613b064	256	if (page == NULL) {
96c77dba	257	xen_pv_printf(&netdev->xendev, 0,
b9730c5b	258	"error: rx gref dereference failed (%d)\n",
e613b064	259	rxreq.gref);
209cd7ab AP	260	net_rx_response(netdev, &rxreq, NETIF_RSP_ERROR, 0, 0, 0);
209cd7ab AP	261	return -1;
e613b064 AL	262	}
e613b064 AL	263	memcpy(page + NET_IP_ALIGN, buf, size);
58560f2a	264	xen_be_unmap_grant_ref(&netdev->xendev, page);
e613b064	265	net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);
4f1c942b MM	266
4f1c942b MM	267	return size;
e613b064 AL	268	}
	269
	270	/* ------------------------------------------------------------- */
	271
658788c5	272	static NetClientInfo net_xen_info = {
f394b2e2	273	.type = NET_CLIENT_DRIVER_NIC,
658788c5	274	.size = sizeof(NICState),
658788c5 MM	275	.receive = net_rx_packet,
	276	};
	277
2d0ed5e6	278	static int net_init(struct XenLegacyDevice *xendev)
e613b064 AL	279	{
e613b064 AL	280	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
e613b064 AL	281
e613b064 AL	282	/* read xenstore entries */
209cd7ab AP	283	if (netdev->mac == NULL) {
	284	netdev->mac = xenstore_read_be_str(&netdev->xendev, "mac");
	285	}
e613b064 AL	286
e613b064 AL	287	/* do we have all we need? */
209cd7ab AP	288	if (netdev->mac == NULL) {
	289	return -1;
	290	}
e613b064	291
209cd7ab	292	if (net_parse_macaddr(netdev->conf.macaddr.a, netdev->mac) < 0) {
658788c5	293	return -1;
209cd7ab	294	}
658788c5	295
658788c5 MM	296	netdev->nic = qemu_new_nic(&net_xen_info, &netdev->conf,
	297	"xen", NULL, netdev);
	298
b356f76d JW	299	snprintf(qemu_get_queue(netdev->nic)->info_str,
b356f76d JW	300	sizeof(qemu_get_queue(netdev->nic)->info_str),
e613b064 AL	301	"nic: xenbus vif macaddr=%s", netdev->mac);
	302
	303	/* fill info */
	304	xenstore_write_be_int(&netdev->xendev, "feature-rx-copy", 1);
	305	xenstore_write_be_int(&netdev->xendev, "feature-rx-flip", 0);
	306
	307	return 0;
	308	}
	309
2d0ed5e6	310	static int net_connect(struct XenLegacyDevice *xendev)
e613b064 AL	311	{
	312	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	313	int rx_copy;
	314
	315	if (xenstore_read_fe_int(&netdev->xendev, "tx-ring-ref",
209cd7ab AP	316	&netdev->tx_ring_ref) == -1) {
	317	return -1;
	318	}
e613b064	319	if (xenstore_read_fe_int(&netdev->xendev, "rx-ring-ref",
209cd7ab AP	320	&netdev->rx_ring_ref) == -1) {
	321	return 1;
	322	}
e613b064	323	if (xenstore_read_fe_int(&netdev->xendev, "event-channel",
209cd7ab AP	324	&netdev->xendev.remote_port) == -1) {
	325	return -1;
	326	}
e613b064	327
209cd7ab AP	328	if (xenstore_read_fe_int(&netdev->xendev, "request-rx-copy", &rx_copy) == -1) {
	329	rx_copy = 0;
	330	}
e613b064	331	if (rx_copy == 0) {
96c77dba	332	xen_pv_printf(&netdev->xendev, 0,
b9730c5b	333	"frontend doesn't support rx-copy.\n");
209cd7ab	334	return -1;
e613b064 AL	335	}
e613b064 AL	336
58560f2a PD	337	netdev->txs = xen_be_map_grant_ref(&netdev->xendev,
	338	netdev->tx_ring_ref,
	339	PROT_READ \| PROT_WRITE);
b4f72e31 CG	340	if (!netdev->txs) {
	341	return -1;
	342	}
58560f2a PD	343	netdev->rxs = xen_be_map_grant_ref(&netdev->xendev,
	344	netdev->rx_ring_ref,
	345	PROT_READ \| PROT_WRITE);
b4f72e31	346	if (!netdev->rxs) {
58560f2a	347	xen_be_unmap_grant_ref(&netdev->xendev, netdev->txs);
b4f72e31	348	netdev->txs = NULL;
209cd7ab AP	349	return -1;
209cd7ab AP	350	}
e613b064 AL	351	BACK_RING_INIT(&netdev->tx_ring, netdev->txs, XC_PAGE_SIZE);
	352	BACK_RING_INIT(&netdev->rx_ring, netdev->rxs, XC_PAGE_SIZE);
	353
	354	xen_be_bind_evtchn(&netdev->xendev);
	355
96c77dba	356	xen_pv_printf(&netdev->xendev, 1, "ok: tx-ring-ref %d, rx-ring-ref %d, "
209cd7ab AP	357	"remote port %d, local port %d\n",
	358	netdev->tx_ring_ref, netdev->rx_ring_ref,
	359	netdev->xendev.remote_port, netdev->xendev.local_port);
3e3cabcf GH	360
3e3cabcf GH	361	net_tx_packets(netdev);
e613b064 AL	362	return 0;
	363	}
	364
2d0ed5e6	365	static void net_disconnect(struct XenLegacyDevice *xendev)
e613b064 AL	366	{
	367	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	368
65807f4b	369	xen_pv_unbind_evtchn(&netdev->xendev);
e613b064 AL	370
e613b064 AL	371	if (netdev->txs) {
58560f2a	372	xen_be_unmap_grant_ref(&netdev->xendev, netdev->txs);
209cd7ab	373	netdev->txs = NULL;
e613b064 AL	374	}
e613b064 AL	375	if (netdev->rxs) {
58560f2a	376	xen_be_unmap_grant_ref(&netdev->xendev, netdev->rxs);
209cd7ab	377	netdev->rxs = NULL;
e613b064	378	}
e613b064 AL	379	}
e613b064 AL	380
2d0ed5e6	381	static void net_event(struct XenLegacyDevice *xendev)
e613b064 AL	382	{
	383	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	384	net_tx_packets(netdev);
b356f76d	385	qemu_flush_queued_packets(qemu_get_queue(netdev->nic));
e613b064 AL	386	}
e613b064 AL	387
2d0ed5e6	388	static int net_free(struct XenLegacyDevice *xendev)
e613b064 AL	389	{
	390	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	391
d4685837 CG	392	if (netdev->nic) {
	393	qemu_del_nic(netdev->nic);
	394	netdev->nic = NULL;
	395	}
7267c094	396	g_free(netdev->mac);
a39d97c7	397	netdev->mac = NULL;
e613b064 AL	398	return 0;
	399	}
	400
	401	/* ------------------------------------------------------------- */
	402
	403	struct XenDevOps xen_netdev_ops = {
	404	.size = sizeof(struct XenNetDev),
	405	.flags = DEVOPS_FLAG_NEED_GNTDEV,
	406	.init = net_init,
384087b2	407	.initialise = net_connect,
e613b064 AL	408	.event = net_event,
	409	.disconnect = net_disconnect,
	410	.free = net_free,
	411	};