[mirror_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 <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <string.h>
#include <unistd.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/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_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(qemu_get_queue(netdev->nic), tmpbuf,
                                 txreq.size);
            } else {
                qemu_send_packet(qemu_get_queue(netdev->nic),
                                 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 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_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),
    .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_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);
    if (!netdev->txs) {
        return -1;
    }
    netdev->rxs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
                                          netdev->xendev.dom,
                                          netdev->rx_ring_ref,
                                          PROT_READ | PROT_WRITE);
    if (!netdev->rxs) {
        xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->txs, 1);
        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_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;
    }
}

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 <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>
e613b064 AL	27	#include <inttypes.h>
	28	#include <fcntl.h>
	29	#include <errno.h>
e613b064 AL	30	#include <sys/socket.h>
	31	#include <sys/ioctl.h>
	32	#include <sys/types.h>
	33	#include <sys/stat.h>
	34	#include <sys/mman.h>
	35	#include <sys/wait.h>
e613b064	36
83c9f4ca	37	#include "hw/hw.h"
1422e32d	38	#include "net/net.h"
7200ac3c	39	#include "net/checksum.h"
658788c5	40	#include "net/util.h"
0d09e41a	41	#include "hw/xen/xen_backend.h"
e613b064	42
b41f6719 AP	43	#include <xen/io/netif.h>
b41f6719 AP	44
e613b064 AL	45	/* ------------------------------------------------------------- */
	46
	47	struct XenNetDev {
	48	struct XenDevice xendev; /* must be first */
	49	char *mac;
	50	int tx_work;
	51	int tx_ring_ref;
	52	int rx_ring_ref;
	53	struct netif_tx_sring *txs;
	54	struct netif_rx_sring *rxs;
	55	netif_tx_back_ring_t tx_ring;
	56	netif_rx_back_ring_t rx_ring;
658788c5 MM	57	NICConf conf;
658788c5 MM	58	NICState *nic;
e613b064 AL	59	};
	60
	61	/* ------------------------------------------------------------- */
	62
	63	static void net_tx_response(struct XenNetDev netdev, netif_tx_request_t txp, int8_t st)
	64	{
	65	RING_IDX i = netdev->tx_ring.rsp_prod_pvt;
	66	netif_tx_response_t *resp;
	67	int notify;
	68
	69	resp = RING_GET_RESPONSE(&netdev->tx_ring, i);
	70	resp->id = txp->id;
	71	resp->status = st;
	72
	73	#if 0
209cd7ab AP	74	if (txp->flags & NETTXF_extra_info) {
	75	RING_GET_RESPONSE(&netdev->tx_ring, ++i)->status = NETIF_RSP_NULL;
	76	}
e613b064 AL	77	#endif
	78
	79	netdev->tx_ring.rsp_prod_pvt = ++i;
	80	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->tx_ring, notify);
209cd7ab AP	81	if (notify) {
	82	xen_be_send_notify(&netdev->xendev);
	83	}
e613b064 AL	84
e613b064 AL	85	if (i == netdev->tx_ring.req_cons) {
209cd7ab AP	86	int more_to_do;
	87	RING_FINAL_CHECK_FOR_REQUESTS(&netdev->tx_ring, more_to_do);
	88	if (more_to_do) {
	89	netdev->tx_work++;
	90	}
e613b064 AL	91	}
	92	}
	93
	94	static void net_tx_error(struct XenNetDev netdev, netif_tx_request_t txp, RING_IDX end)
	95	{
	96	#if 0
	97	/*
	98	* Hmm, why netback fails everything in the ring?
	99	* Should we do that even when not supporting SG and TSO?
	100	*/
	101	RING_IDX cons = netdev->tx_ring.req_cons;
	102
	103	do {
209cd7ab AP	104	make_tx_response(netif, txp, NETIF_RSP_ERROR);
	105	if (cons >= end) {
	106	break;
	107	}
	108	txp = RING_GET_REQUEST(&netdev->tx_ring, cons++);
e613b064 AL	109	} while (1);
	110	netdev->tx_ring.req_cons = cons;
	111	netif_schedule_work(netif);
	112	netif_put(netif);
	113	#else
	114	net_tx_response(netdev, txp, NETIF_RSP_ERROR);
	115	#endif
	116	}
	117
	118	static void net_tx_packets(struct XenNetDev *netdev)
	119	{
	120	netif_tx_request_t txreq;
	121	RING_IDX rc, rp;
	122	void *page;
	123	void *tmpbuf = NULL;
	124
	125	for (;;) {
209cd7ab AP	126	rc = netdev->tx_ring.req_cons;
	127	rp = netdev->tx_ring.sring->req_prod;
	128	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
e613b064	129
209cd7ab AP	130	while ((rc != rp)) {
	131	if (RING_REQUEST_CONS_OVERFLOW(&netdev->tx_ring, rc)) {
	132	break;
	133	}
	134	memcpy(&txreq, RING_GET_REQUEST(&netdev->tx_ring, rc), sizeof(txreq));
	135	netdev->tx_ring.req_cons = ++rc;
e613b064 AL	136
e613b064 AL	137	#if 1
209cd7ab AP	138	/* should not happen in theory, we don't announce the *
	139	* feature-{sg,gso,whatelse} flags in xenstore (yet?) */
	140	if (txreq.flags & NETTXF_extra_info) {
	141	xen_be_printf(&netdev->xendev, 0, "FIXME: extra info flag\n");
	142	net_tx_error(netdev, &txreq, rc);
	143	continue;
	144	}
	145	if (txreq.flags & NETTXF_more_data) {
	146	xen_be_printf(&netdev->xendev, 0, "FIXME: more data flag\n");
	147	net_tx_error(netdev, &txreq, rc);
	148	continue;
	149	}
e613b064 AL	150	#endif
e613b064 AL	151
209cd7ab AP	152	if (txreq.size < 14) {
	153	xen_be_printf(&netdev->xendev, 0, "bad packet size: %d\n", txreq.size);
	154	net_tx_error(netdev, &txreq, rc);
	155	continue;
	156	}
	157
	158	if ((txreq.offset + txreq.size) > XC_PAGE_SIZE) {
	159	xen_be_printf(&netdev->xendev, 0, "error: page crossing\n");
	160	net_tx_error(netdev, &txreq, rc);
	161	continue;
	162	}
	163
	164	xen_be_printf(&netdev->xendev, 3, "tx packet ref %d, off %d, len %d, flags 0x%x%s%s%s%s\n",
	165	txreq.gref, txreq.offset, txreq.size, txreq.flags,
	166	(txreq.flags & NETTXF_csum_blank) ? " csum_blank" : "",
	167	(txreq.flags & NETTXF_data_validated) ? " data_validated" : "",
	168	(txreq.flags & NETTXF_more_data) ? " more_data" : "",
	169	(txreq.flags & NETTXF_extra_info) ? " extra_info" : "");
	170
	171	page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
	172	netdev->xendev.dom,
	173	txreq.gref, PROT_READ);
	174	if (page == NULL) {
	175	xen_be_printf(&netdev->xendev, 0, "error: tx gref dereference failed (%d)\n",
e613b064	176	txreq.gref);
209cd7ab AP	177	net_tx_error(netdev, &txreq, rc);
	178	continue;
	179	}
	180	if (txreq.flags & NETTXF_csum_blank) {
e613b064	181	/* have read-only mapping -> can't fill checksum in-place */
209cd7ab	182	if (!tmpbuf) {
7267c094	183	tmpbuf = g_malloc(XC_PAGE_SIZE);
209cd7ab	184	}
e613b064	185	memcpy(tmpbuf, page + txreq.offset, txreq.size);
209cd7ab	186	net_checksum_calculate(tmpbuf, txreq.size);
b356f76d JW	187	qemu_send_packet(qemu_get_queue(netdev->nic), tmpbuf,
b356f76d JW	188	txreq.size);
e613b064	189	} else {
b356f76d JW	190	qemu_send_packet(qemu_get_queue(netdev->nic),
b356f76d JW	191	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
4e68f7a0	236	static ssize_t net_rx_packet(NetClientState nc, const uint8_t buf, size_t size)
e613b064	237	{
cc1f0f45	238	struct XenNetDev *netdev = qemu_get_nic_opaque(nc);
e613b064 AL	239	netif_rx_request_t rxreq;
	240	RING_IDX rc, rp;
	241	void *page;
	242
209cd7ab AP	243	if (netdev->xendev.be_state != XenbusStateConnected) {
	244	return -1;
	245	}
e613b064 AL	246
	247	rc = netdev->rx_ring.req_cons;
	248	rp = netdev->rx_ring.sring->req_prod;
	249	xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
	250
	251	if (rc == rp \|\| RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
7bba83bf	252	return 0;
e613b064 AL	253	}
e613b064 AL	254	if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
209cd7ab AP	255	xen_be_printf(&netdev->xendev, 0, "packet too big (%lu > %ld)",
	256	(unsigned long)size, XC_PAGE_SIZE - NET_IP_ALIGN);
	257	return -1;
e613b064 AL	258	}
	259
	260	memcpy(&rxreq, RING_GET_REQUEST(&netdev->rx_ring, rc), sizeof(rxreq));
	261	netdev->rx_ring.req_cons = ++rc;
	262
	263	page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	264	netdev->xendev.dom,
209cd7ab AP	265	rxreq.gref, PROT_WRITE);
e613b064	266	if (page == NULL) {
209cd7ab	267	xen_be_printf(&netdev->xendev, 0, "error: rx gref dereference failed (%d)\n",
e613b064	268	rxreq.gref);
209cd7ab AP	269	net_rx_response(netdev, &rxreq, NETIF_RSP_ERROR, 0, 0, 0);
209cd7ab AP	270	return -1;
e613b064 AL	271	}
	272	memcpy(page + NET_IP_ALIGN, buf, size);
	273	xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
	274	net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);
4f1c942b MM	275
4f1c942b MM	276	return size;
e613b064 AL	277	}
	278
	279	/* ------------------------------------------------------------- */
	280
658788c5	281	static NetClientInfo net_xen_info = {
2be64a68	282	.type = NET_CLIENT_OPTIONS_KIND_NIC,
658788c5	283	.size = sizeof(NICState),
658788c5 MM	284	.receive = net_rx_packet,
	285	};
	286
e613b064 AL	287	static int net_init(struct XenDevice *xendev)
	288	{
	289	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
e613b064 AL	290
e613b064 AL	291	/* read xenstore entries */
209cd7ab AP	292	if (netdev->mac == NULL) {
	293	netdev->mac = xenstore_read_be_str(&netdev->xendev, "mac");
	294	}
e613b064 AL	295
e613b064 AL	296	/* do we have all we need? */
209cd7ab AP	297	if (netdev->mac == NULL) {
	298	return -1;
	299	}
e613b064	300
209cd7ab	301	if (net_parse_macaddr(netdev->conf.macaddr.a, netdev->mac) < 0) {
658788c5	302	return -1;
209cd7ab	303	}
658788c5	304
658788c5 MM	305	netdev->nic = qemu_new_nic(&net_xen_info, &netdev->conf,
	306	"xen", NULL, netdev);
	307
b356f76d JW	308	snprintf(qemu_get_queue(netdev->nic)->info_str,
b356f76d JW	309	sizeof(qemu_get_queue(netdev->nic)->info_str),
e613b064 AL	310	"nic: xenbus vif macaddr=%s", netdev->mac);
	311
	312	/* fill info */
	313	xenstore_write_be_int(&netdev->xendev, "feature-rx-copy", 1);
	314	xenstore_write_be_int(&netdev->xendev, "feature-rx-flip", 0);
	315
	316	return 0;
	317	}
	318
	319	static int net_connect(struct XenDevice *xendev)
	320	{
	321	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	322	int rx_copy;
	323
	324	if (xenstore_read_fe_int(&netdev->xendev, "tx-ring-ref",
209cd7ab AP	325	&netdev->tx_ring_ref) == -1) {
	326	return -1;
	327	}
e613b064	328	if (xenstore_read_fe_int(&netdev->xendev, "rx-ring-ref",
209cd7ab AP	329	&netdev->rx_ring_ref) == -1) {
	330	return 1;
	331	}
e613b064	332	if (xenstore_read_fe_int(&netdev->xendev, "event-channel",
209cd7ab AP	333	&netdev->xendev.remote_port) == -1) {
	334	return -1;
	335	}
e613b064	336
209cd7ab AP	337	if (xenstore_read_fe_int(&netdev->xendev, "request-rx-copy", &rx_copy) == -1) {
	338	rx_copy = 0;
	339	}
e613b064	340	if (rx_copy == 0) {
209cd7ab AP	341	xen_be_printf(&netdev->xendev, 0, "frontend doesn't support rx-copy.\n");
209cd7ab AP	342	return -1;
e613b064 AL	343	}
	344
	345	netdev->txs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	346	netdev->xendev.dom,
	347	netdev->tx_ring_ref,
	348	PROT_READ \| PROT_WRITE);
b4f72e31 CG	349	if (!netdev->txs) {
	350	return -1;
	351	}
e613b064	352	netdev->rxs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
209cd7ab AP	353	netdev->xendev.dom,
	354	netdev->rx_ring_ref,
	355	PROT_READ \| PROT_WRITE);
b4f72e31 CG	356	if (!netdev->rxs) {
	357	xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->txs, 1);
	358	netdev->txs = NULL;
209cd7ab AP	359	return -1;
209cd7ab AP	360	}
e613b064 AL	361	BACK_RING_INIT(&netdev->tx_ring, netdev->txs, XC_PAGE_SIZE);
	362	BACK_RING_INIT(&netdev->rx_ring, netdev->rxs, XC_PAGE_SIZE);
	363
	364	xen_be_bind_evtchn(&netdev->xendev);
	365
	366	xen_be_printf(&netdev->xendev, 1, "ok: tx-ring-ref %d, rx-ring-ref %d, "
209cd7ab AP	367	"remote port %d, local port %d\n",
	368	netdev->tx_ring_ref, netdev->rx_ring_ref,
	369	netdev->xendev.remote_port, netdev->xendev.local_port);
3e3cabcf GH	370
3e3cabcf GH	371	net_tx_packets(netdev);
e613b064 AL	372	return 0;
	373	}
	374
	375	static void net_disconnect(struct XenDevice *xendev)
	376	{
	377	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	378
	379	xen_be_unbind_evtchn(&netdev->xendev);
	380
	381	if (netdev->txs) {
209cd7ab AP	382	xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->txs, 1);
209cd7ab AP	383	netdev->txs = NULL;
e613b064 AL	384	}
e613b064 AL	385	if (netdev->rxs) {
209cd7ab AP	386	xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->rxs, 1);
209cd7ab AP	387	netdev->rxs = NULL;
e613b064	388	}
e613b064 AL	389	}
	390
	391	static void net_event(struct XenDevice *xendev)
	392	{
	393	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	394	net_tx_packets(netdev);
b356f76d	395	qemu_flush_queued_packets(qemu_get_queue(netdev->nic));
e613b064 AL	396	}
	397
	398	static int net_free(struct XenDevice *xendev)
	399	{
	400	struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
	401
d4685837 CG	402	if (netdev->nic) {
	403	qemu_del_nic(netdev->nic);
	404	netdev->nic = NULL;
	405	}
7267c094	406	g_free(netdev->mac);
a39d97c7	407	netdev->mac = NULL;
e613b064 AL	408	return 0;
	409	}
	410
	411	/* ------------------------------------------------------------- */
	412
	413	struct XenDevOps xen_netdev_ops = {
	414	.size = sizeof(struct XenNetDev),
	415	.flags = DEVOPS_FLAG_NEED_GNTDEV,
	416	.init = net_init,
384087b2	417	.initialise = net_connect,
e613b064 AL	418	.event = net_event,
	419	.disconnect = net_disconnect,
	420	.free = net_free,
	421	};