[qemu.git] / hw / net / xen_nic.c

/*
 *  xen paravirt network card backend
 *
 *  (c) Gerd Hoffmann <[email protected]>
 *
 *  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 "hw/hw.h"
#include "net/net.h"
#include "net/checksum.h"
#include "net/util.h"
#include "hw/xen/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_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 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->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 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_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 XenDevice *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 XenDevice *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 XenDevice *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 <[email protected]>
	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
83c9f4ca	27	#include "hw/hw.h"
1422e32d	28	#include "net/net.h"
7200ac3c	29	#include "net/checksum.h"
658788c5	30	#include "net/util.h"
0d09e41a	31	#include "hw/xen/xen_backend.h"
e613b064	32
b41f6719 AP	33	#include <xen/io/netif.h>
b41f6719 AP	34
e613b064 AL	35	/* ------------------------------------------------------------- */
	36
	37	struct XenNetDev {
	38	struct XenDevice xendev; /* must be first */
	39	char *mac;
	40	int tx_work;
	41	int tx_ring_ref;
	42	int rx_ring_ref;
	43	struct netif_tx_sring *txs;
	44	struct netif_rx_sring *rxs;
	45	netif_tx_back_ring_t tx_ring;
	46	netif_rx_back_ring_t rx_ring;
658788c5 MM	47	NICConf conf;
658788c5 MM	48	NICState *nic;
e613b064 AL	49	};
	50
	51	/* ------------------------------------------------------------- */
	52
	53	static void net_tx_response(struct XenNetDev netdev, netif_tx_request_t txp, int8_t st)
	54	{
	55	RING_IDX i = netdev->tx_ring.rsp_prod_pvt;
	56	netif_tx_response_t *resp;
	57	int notify;
	58
	59	resp = RING_GET_RESPONSE(&netdev->tx_ring, i);
	60	resp->id = txp->id;
	61	resp->status = st;
	62
	63	#if 0
209cd7ab AP	64	if (txp->flags & NETTXF_extra_info) {
	65	RING_GET_RESPONSE(&netdev->tx_ring, ++i)->status = NETIF_RSP_NULL;
	66	}
e613b064 AL	67	#endif
	68
	69	netdev->tx_ring.rsp_prod_pvt = ++i;
	70	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->tx_ring, notify);
209cd7ab	71	if (notify) {
ba18fa2a	72	xen_pv_send_notify(&netdev->xendev);
209cd7ab	73	}
e613b064 AL	74
e613b064 AL	75	if (i == netdev->tx_ring.req_cons) {
209cd7ab AP	76	int more_to_do;
	77	RING_FINAL_CHECK_FOR_REQUESTS(&netdev->tx_ring, more_to_do);
	78	if (more_to_do) {
	79	netdev->tx_work++;
	80	}
e613b064 AL	81	}
	82	}
	83
	84	static void net_tx_error(struct XenNetDev netdev, netif_tx_request_t txp, RING_IDX end)
	85	{
	86	#if 0
	87	/*
	88	* Hmm, why netback fails everything in the ring?
	89	* Should we do that even when not supporting SG and TSO?
	90	*/
	91	RING_IDX cons = netdev->tx_ring.req_cons;
	92
	93	do {
209cd7ab AP	94	make_tx_response(netif, txp, NETIF_RSP_ERROR);
	95	if (cons >= end) {
	96	break;
	97	}
	98	txp = RING_GET_REQUEST(&netdev->tx_ring, cons++);
e613b064 AL	99	} while (1);
	100	netdev->tx_ring.req_cons = cons;
	101	netif_schedule_work(netif);
	102	netif_put(netif);
	103	#else
	104	net_tx_response(netdev, txp, NETIF_RSP_ERROR);
	105	#endif
	106	}
	107
	108	static void net_tx_packets(struct XenNetDev *netdev)
	109	{
	110	netif_tx_request_t txreq;
	111	RING_IDX rc, rp;
	112	void *page;
	113	void *tmpbuf = NULL;
	114
	115	for (;;) {
209cd7ab AP	116	rc = netdev->tx_ring.req_cons;
	117	rp = netdev->tx_ring.sring->req_prod;
	118	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
e613b064	119
209cd7ab AP	120	while ((rc != rp)) {
	121	if (RING_REQUEST_CONS_OVERFLOW(&netdev->tx_ring, rc)) {
	122	break;
	123	}
	124	memcpy(&txreq, RING_GET_REQUEST(&netdev->tx_ring, rc), sizeof(txreq));
	125	netdev->tx_ring.req_cons = ++rc;
e613b064 AL	126
e613b064 AL	127	#if 1
209cd7ab AP	128	/* should not happen in theory, we don't announce the *
	129	* feature-{sg,gso,whatelse} flags in xenstore (yet?) */
	130	if (txreq.flags & NETTXF_extra_info) {
96c77dba	131	xen_pv_printf(&netdev->xendev, 0, "FIXME: extra info flag\n");
209cd7ab AP	132	net_tx_error(netdev, &txreq, rc);
	133	continue;
	134	}
	135	if (txreq.flags & NETTXF_more_data) {
96c77dba	136	xen_pv_printf(&netdev->xendev, 0, "FIXME: more data flag\n");
209cd7ab AP	137	net_tx_error(netdev, &txreq, rc);
	138	continue;
	139	}
e613b064 AL	140	#endif
e613b064 AL	141
209cd7ab	142	if (txreq.size < 14) {
96c77dba	143	xen_pv_printf(&netdev->xendev, 0, "bad packet size: %d\n",
b9730c5b	144	txreq.size);
209cd7ab AP	145	net_tx_error(netdev, &txreq, rc);
	146	continue;
	147	}
	148
	149	if ((txreq.offset + txreq.size) > XC_PAGE_SIZE) {
96c77dba	150	xen_pv_printf(&netdev->xendev, 0, "error: page crossing\n");
209cd7ab AP	151	net_tx_error(netdev, &txreq, rc);
	152	continue;
	153	}
	154
96c77dba	155	xen_pv_printf(&netdev->xendev, 3,
c22e91b1	156	"tx packet ref %d, off %d, len %d, flags 0x%x%s%s%s%s\n",
209cd7ab AP	157	txreq.gref, txreq.offset, txreq.size, txreq.flags,
	158	(txreq.flags & NETTXF_csum_blank) ? " csum_blank" : "",
	159	(txreq.flags & NETTXF_data_validated) ? " data_validated" : "",
	160	(txreq.flags & NETTXF_more_data) ? " more_data" : "",
	161	(txreq.flags & NETTXF_extra_info) ? " extra_info" : "");
	162
58560f2a PD	163	page = xen_be_map_grant_ref(&netdev->xendev, txreq.gref,
58560f2a PD	164	PROT_READ);
209cd7ab	165	if (page == NULL) {
96c77dba	166	xen_pv_printf(&netdev->xendev, 0,
c22e91b1 EC	167	"error: tx gref dereference failed (%d)\n",
c22e91b1 EC	168	txreq.gref);
209cd7ab AP	169	net_tx_error(netdev, &txreq, rc);
	170	continue;
	171	}
	172	if (txreq.flags & NETTXF_csum_blank) {
e613b064	173	/* have read-only mapping -> can't fill checksum in-place */
209cd7ab	174	if (!tmpbuf) {
7267c094	175	tmpbuf = g_malloc(XC_PAGE_SIZE);
209cd7ab	176	}
e613b064	177	memcpy(tmpbuf, page + txreq.offset, txreq.size);
209cd7ab	178	net_checksum_calculate(tmpbuf, txreq.size);
b356f76d JW	179	qemu_send_packet(qemu_get_queue(netdev->nic), tmpbuf,
b356f76d JW	180	txreq.size);
e613b064	181	} else {
b356f76d JW	182	qemu_send_packet(qemu_get_queue(netdev->nic),
b356f76d JW	183	page + txreq.offset, txreq.size);
e613b064	184	}
58560f2a	185	xen_be_unmap_grant_ref(&netdev->xendev, page);
209cd7ab AP	186	net_tx_response(netdev, &txreq, NETIF_RSP_OKAY);
	187	}
	188	if (!netdev->tx_work) {
	189	break;
	190	}
	191	netdev->tx_work = 0;
e613b064	192	}
7267c094	193	g_free(tmpbuf);
e613b064 AL	194	}
	195
	196	/* ------------------------------------------------------------- */
	197
	198	static void net_rx_response(struct XenNetDev *netdev,
209cd7ab AP	199	netif_rx_request_t *req, int8_t st,
	200	uint16_t offset, uint16_t size,
	201	uint16_t flags)
e613b064 AL	202	{
	203	RING_IDX i = netdev->rx_ring.rsp_prod_pvt;
	204	netif_rx_response_t *resp;
	205	int notify;
	206
	207	resp = RING_GET_RESPONSE(&netdev->rx_ring, i);
	208	resp->offset = offset;
	209	resp->flags = flags;
	210	resp->id = req->id;
	211	resp->status = (int16_t)size;
209cd7ab AP	212	if (st < 0) {
	213	resp->status = (int16_t)st;
	214	}
e613b064	215
96c77dba	216	xen_pv_printf(&netdev->xendev, 3,
b9730c5b	217	"rx response: idx %d, status %d, flags 0x%x\n",
209cd7ab	218	i, resp->status, resp->flags);
e613b064 AL	219
	220	netdev->rx_ring.rsp_prod_pvt = ++i;
	221	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->rx_ring, notify);
209cd7ab	222	if (notify) {
ba18fa2a	223	xen_pv_send_notify(&netdev->xendev);
209cd7ab	224	}
e613b064 AL	225	}
	226
	227	#define NET_IP_ALIGN 2
	228
4e68f7a0	229	static ssize_t net_rx_packet(NetClientState nc, const uint8_t buf, size_t size)
e613b064	230	{
cc1f0f45	231	struct XenNetDev *netdev = qemu_get_nic_opaque(nc);
e613b064 AL	232	netif_rx_request_t rxreq;
	233	RING_IDX rc, rp;
	234	void *page;
	235
209cd7ab AP	236	if (netdev->xendev.be_state != XenbusStateConnected) {
	237	return -1;
	238	}
e613b064 AL	239
	240	rc = netdev->rx_ring.req_cons;
	241	rp = netdev->rx_ring.sring->req_prod;
	242	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
	243
	244	if (rc == rp \|\| RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
7bba83bf	245	return 0;
e613b064 AL	246	}
e613b064 AL	247	if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
96c77dba	248	xen_pv_printf(&netdev->xendev, 0, "packet too big (%lu > %ld)",
209cd7ab AP	249	(unsigned long)size, XC_PAGE_SIZE - NET_IP_ALIGN);
209cd7ab AP	250	return -1;
e613b064 AL	251	}
	252
	253	memcpy(&rxreq, RING_GET_REQUEST(&netdev->rx_ring, rc), sizeof(rxreq));
	254	netdev->rx_ring.req_cons = ++rc;
	255
58560f2a	256	page = xen_be_map_grant_ref(&netdev->xendev, rxreq.gref, PROT_WRITE);
e613b064	257	if (page == NULL) {
96c77dba	258	xen_pv_printf(&netdev->xendev, 0,
b9730c5b	259	"error: rx gref dereference failed (%d)\n",
e613b064	260	rxreq.gref);
209cd7ab AP	261	net_rx_response(netdev, &rxreq, NETIF_RSP_ERROR, 0, 0, 0);
209cd7ab AP	262	return -1;
e613b064 AL	263	}
e613b064 AL	264	memcpy(page + NET_IP_ALIGN, buf, size);
58560f2a	265	xen_be_unmap_grant_ref(&netdev->xendev, page);
e613b064	266	net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);
4f1c942b MM	267
4f1c942b MM	268	return size;
e613b064 AL	269	}
	270
	271	/* ------------------------------------------------------------- */
	272
658788c5	273	static NetClientInfo net_xen_info = {
f394b2e2	274	.type = NET_CLIENT_DRIVER_NIC,
658788c5	275	.size = sizeof(NICState),
658788c5 MM	276	.receive = net_rx_packet,
	277	};
	278
e613b064 AL	279	static int net_init(struct XenDevice *xendev)
	280	{
	281	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
e613b064 AL	282
e613b064 AL	283	/* read xenstore entries */
209cd7ab AP	284	if (netdev->mac == NULL) {
	285	netdev->mac = xenstore_read_be_str(&netdev->xendev, "mac");
	286	}
e613b064 AL	287
e613b064 AL	288	/* do we have all we need? */
209cd7ab AP	289	if (netdev->mac == NULL) {
	290	return -1;
	291	}
e613b064	292
209cd7ab	293	if (net_parse_macaddr(netdev->conf.macaddr.a, netdev->mac) < 0) {
658788c5	294	return -1;
209cd7ab	295	}
658788c5	296
658788c5 MM	297	netdev->nic = qemu_new_nic(&net_xen_info, &netdev->conf,
	298	"xen", NULL, netdev);
	299
b356f76d JW	300	snprintf(qemu_get_queue(netdev->nic)->info_str,
b356f76d JW	301	sizeof(qemu_get_queue(netdev->nic)->info_str),
e613b064 AL	302	"nic: xenbus vif macaddr=%s", netdev->mac);
	303
	304	/* fill info */
	305	xenstore_write_be_int(&netdev->xendev, "feature-rx-copy", 1);
	306	xenstore_write_be_int(&netdev->xendev, "feature-rx-flip", 0);
	307
	308	return 0;
	309	}
	310
	311	static int net_connect(struct XenDevice *xendev)
	312	{
	313	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	314	int rx_copy;
	315
	316	if (xenstore_read_fe_int(&netdev->xendev, "tx-ring-ref",
209cd7ab AP	317	&netdev->tx_ring_ref) == -1) {
	318	return -1;
	319	}
e613b064	320	if (xenstore_read_fe_int(&netdev->xendev, "rx-ring-ref",
209cd7ab AP	321	&netdev->rx_ring_ref) == -1) {
	322	return 1;
	323	}
e613b064	324	if (xenstore_read_fe_int(&netdev->xendev, "event-channel",
209cd7ab AP	325	&netdev->xendev.remote_port) == -1) {
	326	return -1;
	327	}
e613b064	328
209cd7ab AP	329	if (xenstore_read_fe_int(&netdev->xendev, "request-rx-copy", &rx_copy) == -1) {
	330	rx_copy = 0;
	331	}
e613b064	332	if (rx_copy == 0) {
96c77dba	333	xen_pv_printf(&netdev->xendev, 0,
b9730c5b	334	"frontend doesn't support rx-copy.\n");
209cd7ab	335	return -1;
e613b064 AL	336	}
e613b064 AL	337
58560f2a PD	338	netdev->txs = xen_be_map_grant_ref(&netdev->xendev,
	339	netdev->tx_ring_ref,
	340	PROT_READ \| PROT_WRITE);
b4f72e31 CG	341	if (!netdev->txs) {
	342	return -1;
	343	}
58560f2a PD	344	netdev->rxs = xen_be_map_grant_ref(&netdev->xendev,
	345	netdev->rx_ring_ref,
	346	PROT_READ \| PROT_WRITE);
b4f72e31	347	if (!netdev->rxs) {
58560f2a	348	xen_be_unmap_grant_ref(&netdev->xendev, netdev->txs);
b4f72e31	349	netdev->txs = NULL;
209cd7ab AP	350	return -1;
209cd7ab AP	351	}
e613b064 AL	352	BACK_RING_INIT(&netdev->tx_ring, netdev->txs, XC_PAGE_SIZE);
	353	BACK_RING_INIT(&netdev->rx_ring, netdev->rxs, XC_PAGE_SIZE);
	354
	355	xen_be_bind_evtchn(&netdev->xendev);
	356
96c77dba	357	xen_pv_printf(&netdev->xendev, 1, "ok: tx-ring-ref %d, rx-ring-ref %d, "
209cd7ab AP	358	"remote port %d, local port %d\n",
	359	netdev->tx_ring_ref, netdev->rx_ring_ref,
	360	netdev->xendev.remote_port, netdev->xendev.local_port);
3e3cabcf GH	361
3e3cabcf GH	362	net_tx_packets(netdev);
e613b064 AL	363	return 0;
	364	}
	365
	366	static void net_disconnect(struct XenDevice *xendev)
	367	{
	368	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	369
65807f4b	370	xen_pv_unbind_evtchn(&netdev->xendev);
e613b064 AL	371
e613b064 AL	372	if (netdev->txs) {
58560f2a	373	xen_be_unmap_grant_ref(&netdev->xendev, netdev->txs);
209cd7ab	374	netdev->txs = NULL;
e613b064 AL	375	}
e613b064 AL	376	if (netdev->rxs) {
58560f2a	377	xen_be_unmap_grant_ref(&netdev->xendev, netdev->rxs);
209cd7ab	378	netdev->rxs = NULL;
e613b064	379	}
e613b064 AL	380	}
	381
	382	static void net_event(struct XenDevice *xendev)
	383	{
	384	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	385	net_tx_packets(netdev);
b356f76d	386	qemu_flush_queued_packets(qemu_get_queue(netdev->nic));
e613b064 AL	387	}
	388
	389	static int net_free(struct XenDevice *xendev)
	390	{
	391	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	392
d4685837 CG	393	if (netdev->nic) {
	394	qemu_del_nic(netdev->nic);
	395	netdev->nic = NULL;
	396	}
7267c094	397	g_free(netdev->mac);
a39d97c7	398	netdev->mac = NULL;
e613b064 AL	399	return 0;
	400	}
	401
	402	/* ------------------------------------------------------------- */
	403
	404	struct XenDevOps xen_netdev_ops = {
	405	.size = sizeof(struct XenNetDev),
	406	.flags = DEVOPS_FLAG_NEED_GNTDEV,
	407	.init = net_init,
384087b2	408	.initialise = net_connect,
e613b064 AL	409	.event = net_event,
	410	.disconnect = net_disconnect,
	411	.free = net_free,
	412	};