[qemu.git] / hw / 9pfs / xen-9p-backend.c

/*
 * Xen 9p backend
 *
 * Copyright Aporeto 2017
 *
 * Authors:
 *  Stefano Stabellini <[email protected]>
 *
 */

#include "qemu/osdep.h"

#include "hw/hw.h"
#include "hw/9pfs/9p.h"
#include "hw/xen/xen_backend.h"
#include "hw/9pfs/xen-9pfs.h"
#include "qemu/config-file.h"
#include "fsdev/qemu-fsdev.h"

#define VERSIONS "1"
#define MAX_RINGS 8
#define MAX_RING_ORDER 8

typedef struct Xen9pfsRing {
    struct Xen9pfsDev *priv;

    int ref;
    xenevtchn_handle   *evtchndev;
    int evtchn;
    int local_port;
    int ring_order;
    struct xen_9pfs_data_intf *intf;
    unsigned char *data;
    struct xen_9pfs_data ring;

    struct iovec *sg;
    QEMUBH *bh;

    /* local copies, so that we can read/write PDU data directly from
     * the ring */
    RING_IDX out_cons, out_size, in_cons;
    bool inprogress;
} Xen9pfsRing;

typedef struct Xen9pfsDev {
    struct XenDevice xendev;  /* must be first */
    V9fsState state;
    char *path;
    char *security_model;
    char *tag;
    char *id;

    int num_rings;
    Xen9pfsRing *rings;
} Xen9pfsDev;

static void xen_9pfs_disconnect(struct XenDevice *xendev);

static void xen_9pfs_in_sg(Xen9pfsRing *ring,
                           struct iovec *in_sg,
                           int *num,
                           uint32_t idx,
                           uint32_t size)
{
    RING_IDX cons, prod, masked_prod, masked_cons;

    cons = ring->intf->in_cons;
    prod = ring->intf->in_prod;
    xen_rmb();
    masked_prod = xen_9pfs_mask(prod, XEN_FLEX_RING_SIZE(ring->ring_order));
    masked_cons = xen_9pfs_mask(cons, XEN_FLEX_RING_SIZE(ring->ring_order));

    if (masked_prod < masked_cons) {
        in_sg[0].iov_base = ring->ring.in + masked_prod;
        in_sg[0].iov_len = masked_cons - masked_prod;
        *num = 1;
    } else {
        in_sg[0].iov_base = ring->ring.in + masked_prod;
        in_sg[0].iov_len = XEN_FLEX_RING_SIZE(ring->ring_order) - masked_prod;
        in_sg[1].iov_base = ring->ring.in;
        in_sg[1].iov_len = masked_cons;
        *num = 2;
    }
}

static void xen_9pfs_out_sg(Xen9pfsRing *ring,
                            struct iovec *out_sg,
                            int *num,
                            uint32_t idx)
{
    RING_IDX cons, prod, masked_prod, masked_cons;

    cons = ring->intf->out_cons;
    prod = ring->intf->out_prod;
    xen_rmb();
    masked_prod = xen_9pfs_mask(prod, XEN_FLEX_RING_SIZE(ring->ring_order));
    masked_cons = xen_9pfs_mask(cons, XEN_FLEX_RING_SIZE(ring->ring_order));

    if (masked_cons < masked_prod) {
        out_sg[0].iov_base = ring->ring.out + masked_cons;
        out_sg[0].iov_len = ring->out_size;
        *num = 1;
    } else {
        if (ring->out_size >
            (XEN_FLEX_RING_SIZE(ring->ring_order) - masked_cons)) {
            out_sg[0].iov_base = ring->ring.out + masked_cons;
            out_sg[0].iov_len = XEN_FLEX_RING_SIZE(ring->ring_order) -
                                masked_cons;
            out_sg[1].iov_base = ring->ring.out;
            out_sg[1].iov_len = ring->out_size -
                                (XEN_FLEX_RING_SIZE(ring->ring_order) -
                                 masked_cons);
            *num = 2;
        } else {
            out_sg[0].iov_base = ring->ring.out + masked_cons;
            out_sg[0].iov_len = ring->out_size;
            *num = 1;
        }
    }
}

static ssize_t xen_9pfs_pdu_vmarshal(V9fsPDU *pdu,
                                     size_t offset,
                                     const char *fmt,
                                     va_list ap)
{
    Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
    struct iovec in_sg[2];
    int num;
    ssize_t ret;

    xen_9pfs_in_sg(&xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings],
                   in_sg, &num, pdu->idx, ROUND_UP(offset + 128, 512));

    ret = v9fs_iov_vmarshal(in_sg, num, offset, 0, fmt, ap);
    if (ret < 0) {
        xen_pv_printf(&xen_9pfs->xendev, 0,
                      "Failed to encode VirtFS request type %d\n", pdu->id + 1);
        xen_be_set_state(&xen_9pfs->xendev, XenbusStateClosing);
        xen_9pfs_disconnect(&xen_9pfs->xendev);
    }
    return ret;
}

static ssize_t xen_9pfs_pdu_vunmarshal(V9fsPDU *pdu,
                                       size_t offset,
                                       const char *fmt,
                                       va_list ap)
{
    Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
    struct iovec out_sg[2];
    int num;
    ssize_t ret;

    xen_9pfs_out_sg(&xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings],
                    out_sg, &num, pdu->idx);

    ret = v9fs_iov_vunmarshal(out_sg, num, offset, 0, fmt, ap);
    if (ret < 0) {
        xen_pv_printf(&xen_9pfs->xendev, 0,
                      "Failed to decode VirtFS request type %d\n", pdu->id);
        xen_be_set_state(&xen_9pfs->xendev, XenbusStateClosing);
        xen_9pfs_disconnect(&xen_9pfs->xendev);
    }
    return ret;
}

static void xen_9pfs_init_out_iov_from_pdu(V9fsPDU *pdu,
                                           struct iovec **piov,
                                           unsigned int *pniov,
                                           size_t size)
{
    Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
    Xen9pfsRing *ring = &xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings];
    int num;

    g_free(ring->sg);

    ring->sg = g_malloc0(sizeof(*ring->sg) * 2);
    xen_9pfs_out_sg(ring, ring->sg, &num, pdu->idx);
    *piov = ring->sg;
    *pniov = num;
}

static void xen_9pfs_init_in_iov_from_pdu(V9fsPDU *pdu,
                                          struct iovec **piov,
                                          unsigned int *pniov,
                                          size_t size)
{
    Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
    Xen9pfsRing *ring = &xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings];
    int num;
    size_t buf_size;

    g_free(ring->sg);

    ring->sg = g_malloc0(sizeof(*ring->sg) * 2);
    xen_9pfs_in_sg(ring, ring->sg, &num, pdu->idx, size);

    buf_size = iov_size(ring->sg, num);
    if (buf_size  < size) {
        xen_pv_printf(&xen_9pfs->xendev, 0, "Xen 9pfs request type %d"
                "needs %zu bytes, buffer has %zu\n", pdu->id, size,
                buf_size);
        xen_be_set_state(&xen_9pfs->xendev, XenbusStateClosing);
        xen_9pfs_disconnect(&xen_9pfs->xendev);
    }

    *piov = ring->sg;
    *pniov = num;
}

static void xen_9pfs_push_and_notify(V9fsPDU *pdu)
{
    RING_IDX prod;
    Xen9pfsDev *priv = container_of(pdu->s, Xen9pfsDev, state);
    Xen9pfsRing *ring = &priv->rings[pdu->tag % priv->num_rings];

    g_free(ring->sg);
    ring->sg = NULL;

    ring->intf->out_cons = ring->out_cons;
    xen_wmb();

    prod = ring->intf->in_prod;
    xen_rmb();
    ring->intf->in_prod = prod + pdu->size;
    xen_wmb();

    ring->inprogress = false;
    xenevtchn_notify(ring->evtchndev, ring->local_port);

    qemu_bh_schedule(ring->bh);
}

static const struct V9fsTransport xen_9p_transport = {
    .pdu_vmarshal = xen_9pfs_pdu_vmarshal,
    .pdu_vunmarshal = xen_9pfs_pdu_vunmarshal,
    .init_in_iov_from_pdu = xen_9pfs_init_in_iov_from_pdu,
    .init_out_iov_from_pdu = xen_9pfs_init_out_iov_from_pdu,
    .push_and_notify = xen_9pfs_push_and_notify,
};

static int xen_9pfs_init(struct XenDevice *xendev)
{
    return 0;
}

static int xen_9pfs_receive(Xen9pfsRing *ring)
{
    P9MsgHeader h;
    RING_IDX cons, prod, masked_prod, masked_cons, queued;
    V9fsPDU *pdu;

    if (ring->inprogress) {
        return 0;
    }

    cons = ring->intf->out_cons;
    prod = ring->intf->out_prod;
    xen_rmb();

    queued = xen_9pfs_queued(prod, cons, XEN_FLEX_RING_SIZE(ring->ring_order));
    if (queued < sizeof(h)) {
        return 0;
    }
    ring->inprogress = true;

    masked_prod = xen_9pfs_mask(prod, XEN_FLEX_RING_SIZE(ring->ring_order));
    masked_cons = xen_9pfs_mask(cons, XEN_FLEX_RING_SIZE(ring->ring_order));

    xen_9pfs_read_packet((uint8_t *) &h, ring->ring.out, sizeof(h),
                         masked_prod, &masked_cons,
                         XEN_FLEX_RING_SIZE(ring->ring_order));
    if (queued < le32_to_cpu(h.size_le)) {
        return 0;
    }

    /* cannot fail, because we only handle one request per ring at a time */
    pdu = pdu_alloc(&ring->priv->state);
    ring->out_size = le32_to_cpu(h.size_le);
    ring->out_cons = cons + le32_to_cpu(h.size_le);

    pdu_submit(pdu, &h);

    return 0;
}

static void xen_9pfs_bh(void *opaque)
{
    Xen9pfsRing *ring = opaque;
    xen_9pfs_receive(ring);
}

static void xen_9pfs_evtchn_event(void *opaque)
{
    Xen9pfsRing *ring = opaque;
    evtchn_port_t port;

    port = xenevtchn_pending(ring->evtchndev);
    xenevtchn_unmask(ring->evtchndev, port);

    qemu_bh_schedule(ring->bh);
}

static void xen_9pfs_disconnect(struct XenDevice *xendev)
{
    Xen9pfsDev *xen_9pdev = container_of(xendev, Xen9pfsDev, xendev);
    int i;

    for (i = 0; i < xen_9pdev->num_rings; i++) {
        if (xen_9pdev->rings[i].evtchndev != NULL) {
            qemu_set_fd_handler(xenevtchn_fd(xen_9pdev->rings[i].evtchndev),
                    NULL, NULL, NULL);
            xenevtchn_unbind(xen_9pdev->rings[i].evtchndev,
                             xen_9pdev->rings[i].local_port);
            xen_9pdev->rings[i].evtchndev = NULL;
        }
    }
}

static int xen_9pfs_free(struct XenDevice *xendev)
{
    Xen9pfsDev *xen_9pdev = container_of(xendev, Xen9pfsDev, xendev);
    int i;

    if (xen_9pdev->rings[0].evtchndev != NULL) {
        xen_9pfs_disconnect(xendev);
    }

    for (i = 0; i < xen_9pdev->num_rings; i++) {
        if (xen_9pdev->rings[i].data != NULL) {
            xengnttab_unmap(xen_9pdev->xendev.gnttabdev,
                    xen_9pdev->rings[i].data,
                    (1 << xen_9pdev->rings[i].ring_order));
        }
        if (xen_9pdev->rings[i].intf != NULL) {
            xengnttab_unmap(xen_9pdev->xendev.gnttabdev,
                    xen_9pdev->rings[i].intf,
                    1);
        }
        if (xen_9pdev->rings[i].bh != NULL) {
            qemu_bh_delete(xen_9pdev->rings[i].bh);
        }
    }

    g_free(xen_9pdev->id);
    g_free(xen_9pdev->tag);
    g_free(xen_9pdev->path);
    g_free(xen_9pdev->security_model);
    g_free(xen_9pdev->rings);
    return 0;
}

static int xen_9pfs_connect(struct XenDevice *xendev)
{
    int i;
    Xen9pfsDev *xen_9pdev = container_of(xendev, Xen9pfsDev, xendev);
    V9fsState *s = &xen_9pdev->state;
    QemuOpts *fsdev;

    if (xenstore_read_fe_int(&xen_9pdev->xendev, "num-rings",
                             &xen_9pdev->num_rings) == -1 ||
        xen_9pdev->num_rings > MAX_RINGS || xen_9pdev->num_rings < 1) {
        return -1;
    }

    xen_9pdev->rings = g_malloc0(xen_9pdev->num_rings * sizeof(Xen9pfsRing));
    for (i = 0; i < xen_9pdev->num_rings; i++) {
        char *str;
        int ring_order;

        xen_9pdev->rings[i].priv = xen_9pdev;
        xen_9pdev->rings[i].evtchn = -1;
        xen_9pdev->rings[i].local_port = -1;

        str = g_strdup_printf("ring-ref%u", i);
        if (xenstore_read_fe_int(&xen_9pdev->xendev, str,
                                 &xen_9pdev->rings[i].ref) == -1) {
            g_free(str);
            goto out;
        }
        g_free(str);
        str = g_strdup_printf("event-channel-%u", i);
        if (xenstore_read_fe_int(&xen_9pdev->xendev, str,
                                 &xen_9pdev->rings[i].evtchn) == -1) {
            g_free(str);
            goto out;
        }
        g_free(str);

        xen_9pdev->rings[i].intf =  xengnttab_map_grant_ref(
                xen_9pdev->xendev.gnttabdev,
                xen_9pdev->xendev.dom,
                xen_9pdev->rings[i].ref,
                PROT_READ | PROT_WRITE);
        if (!xen_9pdev->rings[i].intf) {
            goto out;
        }
        ring_order = xen_9pdev->rings[i].intf->ring_order;
        if (ring_order > MAX_RING_ORDER) {
            goto out;
        }
        xen_9pdev->rings[i].ring_order = ring_order;
        xen_9pdev->rings[i].data = xengnttab_map_domain_grant_refs(
                xen_9pdev->xendev.gnttabdev,
                (1 << ring_order),
                xen_9pdev->xendev.dom,
                xen_9pdev->rings[i].intf->ref,
                PROT_READ | PROT_WRITE);
        if (!xen_9pdev->rings[i].data) {
            goto out;
        }
        xen_9pdev->rings[i].ring.in = xen_9pdev->rings[i].data;
        xen_9pdev->rings[i].ring.out = xen_9pdev->rings[i].data +
                                       XEN_FLEX_RING_SIZE(ring_order);

        xen_9pdev->rings[i].bh = qemu_bh_new(xen_9pfs_bh, &xen_9pdev->rings[i]);
        xen_9pdev->rings[i].out_cons = 0;
        xen_9pdev->rings[i].out_size = 0;
        xen_9pdev->rings[i].inprogress = false;


        xen_9pdev->rings[i].evtchndev = xenevtchn_open(NULL, 0);
        if (xen_9pdev->rings[i].evtchndev == NULL) {
            goto out;
        }
        qemu_set_cloexec(xenevtchn_fd(xen_9pdev->rings[i].evtchndev));
        xen_9pdev->rings[i].local_port = xenevtchn_bind_interdomain
                                            (xen_9pdev->rings[i].evtchndev,
                                             xendev->dom,
                                             xen_9pdev->rings[i].evtchn);
        if (xen_9pdev->rings[i].local_port == -1) {
            xen_pv_printf(xendev, 0,
                          "xenevtchn_bind_interdomain failed port=%d\n",
                          xen_9pdev->rings[i].evtchn);
            goto out;
        }
        xen_pv_printf(xendev, 2, "bind evtchn port %d\n", xendev->local_port);
        qemu_set_fd_handler(xenevtchn_fd(xen_9pdev->rings[i].evtchndev),
                xen_9pfs_evtchn_event, NULL, &xen_9pdev->rings[i]);
    }

    xen_9pdev->security_model = xenstore_read_be_str(xendev, "security_model");
    xen_9pdev->path = xenstore_read_be_str(xendev, "path");
    xen_9pdev->id = s->fsconf.fsdev_id =
        g_strdup_printf("xen9p%d", xendev->dev);
    xen_9pdev->tag = s->fsconf.tag = xenstore_read_fe_str(xendev, "tag");
    v9fs_register_transport(s, &xen_9p_transport);
    fsdev = qemu_opts_create(qemu_find_opts("fsdev"),
            s->fsconf.tag,
            1, NULL);
    qemu_opt_set(fsdev, "fsdriver", "local", NULL);
    qemu_opt_set(fsdev, "path", xen_9pdev->path, NULL);
    qemu_opt_set(fsdev, "security_model", xen_9pdev->security_model, NULL);
    qemu_opts_set_id(fsdev, s->fsconf.fsdev_id);
    qemu_fsdev_add(fsdev);
    v9fs_device_realize_common(s, NULL);

    return 0;

out:
    xen_9pfs_free(xendev);
    return -1;
}

static void xen_9pfs_alloc(struct XenDevice *xendev)
{
    xenstore_write_be_str(xendev, "versions", VERSIONS);
    xenstore_write_be_int(xendev, "max-rings", MAX_RINGS);
    xenstore_write_be_int(xendev, "max-ring-page-order", MAX_RING_ORDER);
}

struct XenDevOps xen_9pfs_ops = {
    .size       = sizeof(Xen9pfsDev),
    .flags      = DEVOPS_FLAG_NEED_GNTDEV,
    .alloc      = xen_9pfs_alloc,
    .init       = xen_9pfs_init,
    .initialise = xen_9pfs_connect,
    .disconnect = xen_9pfs_disconnect,
    .free       = xen_9pfs_free,
};
Commit	Line	Data
b37eeb02 SS	1	/*
	2	* Xen 9p backend
	3	*
	4	* Copyright Aporeto 2017
	5	*
	6	* Authors:
	7	* Stefano Stabellini <[email protected]>
	8	*
	9	*/
	10
	11	#include "qemu/osdep.h"
	12
	13	#include "hw/hw.h"
	14	#include "hw/9pfs/9p.h"
	15	#include "hw/xen/xen_backend.h"
	16	#include "hw/9pfs/xen-9pfs.h"
	17	#include "qemu/config-file.h"
	18	#include "fsdev/qemu-fsdev.h"
	19
f23ef34a SS	20	#define VERSIONS "1"
	21	#define MAX_RINGS 8
	22	#define MAX_RING_ORDER 8
	23
	24	typedef struct Xen9pfsRing {
	25	struct Xen9pfsDev *priv;
	26
	27	int ref;
	28	xenevtchn_handle *evtchndev;
	29	int evtchn;
	30	int local_port;
	31	int ring_order;
	32	struct xen_9pfs_data_intf *intf;
	33	unsigned char *data;
	34	struct xen_9pfs_data ring;
	35
	36	struct iovec *sg;
	37	QEMUBH *bh;
	38
	39	/* local copies, so that we can read/write PDU data directly from
	40	* the ring */
	41	RING_IDX out_cons, out_size, in_cons;
	42	bool inprogress;
	43	} Xen9pfsRing;
	44
b37eeb02 SS	45	typedef struct Xen9pfsDev {
b37eeb02 SS	46	struct XenDevice xendev; /* must be first */
f23ef34a SS	47	V9fsState state;
	48	char *path;
	49	char *security_model;
	50	char *tag;
	51	char *id;
	52
	53	int num_rings;
	54	Xen9pfsRing *rings;
b37eeb02 SS	55	} Xen9pfsDev;
b37eeb02 SS	56
e08d1e11 SS	57	static void xen_9pfs_disconnect(struct XenDevice *xendev);
e08d1e11 SS	58
40a23892 SS	59	static void xen_9pfs_in_sg(Xen9pfsRing *ring,
	60	struct iovec *in_sg,
	61	int *num,
	62	uint32_t idx,
	63	uint32_t size)
	64	{
	65	RING_IDX cons, prod, masked_prod, masked_cons;
	66
	67	cons = ring->intf->in_cons;
	68	prod = ring->intf->in_prod;
	69	xen_rmb();
	70	masked_prod = xen_9pfs_mask(prod, XEN_FLEX_RING_SIZE(ring->ring_order));
	71	masked_cons = xen_9pfs_mask(cons, XEN_FLEX_RING_SIZE(ring->ring_order));
	72
	73	if (masked_prod < masked_cons) {
	74	in_sg[0].iov_base = ring->ring.in + masked_prod;
	75	in_sg[0].iov_len = masked_cons - masked_prod;
	76	*num = 1;
	77	} else {
	78	in_sg[0].iov_base = ring->ring.in + masked_prod;
	79	in_sg[0].iov_len = XEN_FLEX_RING_SIZE(ring->ring_order) - masked_prod;
	80	in_sg[1].iov_base = ring->ring.in;
	81	in_sg[1].iov_len = masked_cons;
	82	*num = 2;
	83	}
	84	}
	85
	86	static void xen_9pfs_out_sg(Xen9pfsRing *ring,
	87	struct iovec *out_sg,
	88	int *num,
	89	uint32_t idx)
	90	{
	91	RING_IDX cons, prod, masked_prod, masked_cons;
	92
	93	cons = ring->intf->out_cons;
	94	prod = ring->intf->out_prod;
	95	xen_rmb();
	96	masked_prod = xen_9pfs_mask(prod, XEN_FLEX_RING_SIZE(ring->ring_order));
	97	masked_cons = xen_9pfs_mask(cons, XEN_FLEX_RING_SIZE(ring->ring_order));
	98
	99	if (masked_cons < masked_prod) {
	100	out_sg[0].iov_base = ring->ring.out + masked_cons;
	101	out_sg[0].iov_len = ring->out_size;
	102	*num = 1;
	103	} else {
	104	if (ring->out_size >
	105	(XEN_FLEX_RING_SIZE(ring->ring_order) - masked_cons)) {
	106	out_sg[0].iov_base = ring->ring.out + masked_cons;
	107	out_sg[0].iov_len = XEN_FLEX_RING_SIZE(ring->ring_order) -
	108	masked_cons;
	109	out_sg[1].iov_base = ring->ring.out;
	110	out_sg[1].iov_len = ring->out_size -
	111	(XEN_FLEX_RING_SIZE(ring->ring_order) -
	112	masked_cons);
	113	*num = 2;
	114	} else {
	115	out_sg[0].iov_base = ring->ring.out + masked_cons;
	116	out_sg[0].iov_len = ring->out_size;
	117	*num = 1;
	118	}
	119	}
	120	}
	121
b37eeb02 SS	122	static ssize_t xen_9pfs_pdu_vmarshal(V9fsPDU *pdu,
	123	size_t offset,
	124	const char *fmt,
	125	va_list ap)
	126	{
40a23892 SS	127	Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
	128	struct iovec in_sg[2];
	129	int num;
e08d1e11	130	ssize_t ret;
40a23892 SS	131
	132	xen_9pfs_in_sg(&xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings],
	133	in_sg, &num, pdu->idx, ROUND_UP(offset + 128, 512));
e08d1e11 SS	134
	135	ret = v9fs_iov_vmarshal(in_sg, num, offset, 0, fmt, ap);
	136	if (ret < 0) {
	137	xen_pv_printf(&xen_9pfs->xendev, 0,
	138	"Failed to encode VirtFS request type %d\n", pdu->id + 1);
	139	xen_be_set_state(&xen_9pfs->xendev, XenbusStateClosing);
	140	xen_9pfs_disconnect(&xen_9pfs->xendev);
	141	}
	142	return ret;
b37eeb02 SS	143	}
	144
	145	static ssize_t xen_9pfs_pdu_vunmarshal(V9fsPDU *pdu,
	146	size_t offset,
	147	const char *fmt,
	148	va_list ap)
	149	{
40a23892 SS	150	Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
	151	struct iovec out_sg[2];
	152	int num;
e08d1e11	153	ssize_t ret;
40a23892 SS	154
	155	xen_9pfs_out_sg(&xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings],
	156	out_sg, &num, pdu->idx);
e08d1e11 SS	157
	158	ret = v9fs_iov_vunmarshal(out_sg, num, offset, 0, fmt, ap);
	159	if (ret < 0) {
	160	xen_pv_printf(&xen_9pfs->xendev, 0,
	161	"Failed to decode VirtFS request type %d\n", pdu->id);
	162	xen_be_set_state(&xen_9pfs->xendev, XenbusStateClosing);
	163	xen_9pfs_disconnect(&xen_9pfs->xendev);
	164	}
	165	return ret;
b37eeb02 SS	166	}
	167
	168	static void xen_9pfs_init_out_iov_from_pdu(V9fsPDU *pdu,
	169	struct iovec **piov,
8d37de41 GK	170	unsigned int *pniov,
8d37de41 GK	171	size_t size)
b37eeb02	172	{
40a23892 SS	173	Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
	174	Xen9pfsRing *ring = &xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings];
	175	int num;
	176
	177	g_free(ring->sg);
	178
	179	ring->sg = g_malloc0(sizeof(ring->sg) 2);
	180	xen_9pfs_out_sg(ring, ring->sg, &num, pdu->idx);
	181	*piov = ring->sg;
	182	*pniov = num;
b37eeb02 SS	183	}
	184
	185	static void xen_9pfs_init_in_iov_from_pdu(V9fsPDU *pdu,
	186	struct iovec **piov,
	187	unsigned int *pniov,
	188	size_t size)
	189	{
40a23892 SS	190	Xen9pfsDev *xen_9pfs = container_of(pdu->s, Xen9pfsDev, state);
	191	Xen9pfsRing *ring = &xen_9pfs->rings[pdu->tag % xen_9pfs->num_rings];
	192	int num;
e08d1e11	193	size_t buf_size;
40a23892 SS	194
	195	g_free(ring->sg);
	196
	197	ring->sg = g_malloc0(sizeof(ring->sg) 2);
	198	xen_9pfs_in_sg(ring, ring->sg, &num, pdu->idx, size);
e08d1e11 SS	199
	200	buf_size = iov_size(ring->sg, num);
	201	if (buf_size < size) {
	202	xen_pv_printf(&xen_9pfs->xendev, 0, "Xen 9pfs request type %d"
	203	"needs %zu bytes, buffer has %zu\n", pdu->id, size,
	204	buf_size);
	205	xen_be_set_state(&xen_9pfs->xendev, XenbusStateClosing);
	206	xen_9pfs_disconnect(&xen_9pfs->xendev);
	207	}
	208
40a23892 SS	209	*piov = ring->sg;
40a23892 SS	210	*pniov = num;
b37eeb02 SS	211	}
	212
	213	static void xen_9pfs_push_and_notify(V9fsPDU *pdu)
	214	{
4476e09e SS	215	RING_IDX prod;
	216	Xen9pfsDev *priv = container_of(pdu->s, Xen9pfsDev, state);
	217	Xen9pfsRing *ring = &priv->rings[pdu->tag % priv->num_rings];
	218
	219	g_free(ring->sg);
	220	ring->sg = NULL;
	221
	222	ring->intf->out_cons = ring->out_cons;
	223	xen_wmb();
	224
	225	prod = ring->intf->in_prod;
	226	xen_rmb();
	227	ring->intf->in_prod = prod + pdu->size;
	228	xen_wmb();
	229
	230	ring->inprogress = false;
	231	xenevtchn_notify(ring->evtchndev, ring->local_port);
	232
	233	qemu_bh_schedule(ring->bh);
b37eeb02 SS	234	}
	235
	236	static const struct V9fsTransport xen_9p_transport = {
	237	.pdu_vmarshal = xen_9pfs_pdu_vmarshal,
	238	.pdu_vunmarshal = xen_9pfs_pdu_vunmarshal,
	239	.init_in_iov_from_pdu = xen_9pfs_init_in_iov_from_pdu,
	240	.init_out_iov_from_pdu = xen_9pfs_init_out_iov_from_pdu,
	241	.push_and_notify = xen_9pfs_push_and_notify,
	242	};
	243
	244	static int xen_9pfs_init(struct XenDevice *xendev)
	245	{
	246	return 0;
	247	}
	248
47b70fb1 SS	249	static int xen_9pfs_receive(Xen9pfsRing *ring)
	250	{
	251	P9MsgHeader h;
e08d1e11	252	RING_IDX cons, prod, masked_prod, masked_cons, queued;
47b70fb1 SS	253	V9fsPDU *pdu;
	254
	255	if (ring->inprogress) {
	256	return 0;
	257	}
	258
	259	cons = ring->intf->out_cons;
	260	prod = ring->intf->out_prod;
	261	xen_rmb();
	262
e08d1e11 SS	263	queued = xen_9pfs_queued(prod, cons, XEN_FLEX_RING_SIZE(ring->ring_order));
e08d1e11 SS	264	if (queued < sizeof(h)) {
47b70fb1 SS	265	return 0;
	266	}
	267	ring->inprogress = true;
	268
	269	masked_prod = xen_9pfs_mask(prod, XEN_FLEX_RING_SIZE(ring->ring_order));
	270	masked_cons = xen_9pfs_mask(cons, XEN_FLEX_RING_SIZE(ring->ring_order));
	271
	272	xen_9pfs_read_packet((uint8_t *) &h, ring->ring.out, sizeof(h),
	273	masked_prod, &masked_cons,
	274	XEN_FLEX_RING_SIZE(ring->ring_order));
e08d1e11 SS	275	if (queued < le32_to_cpu(h.size_le)) {
	276	return 0;
	277	}
47b70fb1 SS	278
	279	/* cannot fail, because we only handle one request per ring at a time */
	280	pdu = pdu_alloc(&ring->priv->state);
47b70fb1 SS	281	ring->out_size = le32_to_cpu(h.size_le);
	282	ring->out_cons = cons + le32_to_cpu(h.size_le);
	283
506f3275	284	pdu_submit(pdu, &h);
47b70fb1 SS	285
	286	return 0;
	287	}
	288
f23ef34a SS	289	static void xen_9pfs_bh(void *opaque)
f23ef34a SS	290	{
47b70fb1 SS	291	Xen9pfsRing *ring = opaque;
47b70fb1 SS	292	xen_9pfs_receive(ring);
f23ef34a SS	293	}
	294
	295	static void xen_9pfs_evtchn_event(void *opaque)
	296	{
47b70fb1 SS	297	Xen9pfsRing *ring = opaque;
	298	evtchn_port_t port;
	299
	300	port = xenevtchn_pending(ring->evtchndev);
	301	xenevtchn_unmask(ring->evtchndev, port);
	302
	303	qemu_bh_schedule(ring->bh);
f23ef34a SS	304	}
f23ef34a SS	305
e08d1e11	306	static void xen_9pfs_disconnect(struct XenDevice *xendev)
b37eeb02	307	{
e08d1e11	308	Xen9pfsDev *xen_9pdev = container_of(xendev, Xen9pfsDev, xendev);
f23ef34a	309	int i;
e08d1e11 SS	310
	311	for (i = 0; i < xen_9pdev->num_rings; i++) {
	312	if (xen_9pdev->rings[i].evtchndev != NULL) {
	313	qemu_set_fd_handler(xenevtchn_fd(xen_9pdev->rings[i].evtchndev),
	314	NULL, NULL, NULL);
	315	xenevtchn_unbind(xen_9pdev->rings[i].evtchndev,
	316	xen_9pdev->rings[i].local_port);
	317	xen_9pdev->rings[i].evtchndev = NULL;
	318	}
	319	}
	320	}
	321
	322	static int xen_9pfs_free(struct XenDevice *xendev)
	323	{
f23ef34a	324	Xen9pfsDev *xen_9pdev = container_of(xendev, Xen9pfsDev, xendev);
e08d1e11	325	int i;
f23ef34a	326
e08d1e11 SS	327	if (xen_9pdev->rings[0].evtchndev != NULL) {
	328	xen_9pfs_disconnect(xendev);
	329	}
f23ef34a SS	330
	331	for (i = 0; i < xen_9pdev->num_rings; i++) {
	332	if (xen_9pdev->rings[i].data != NULL) {
	333	xengnttab_unmap(xen_9pdev->xendev.gnttabdev,
	334	xen_9pdev->rings[i].data,
	335	(1 << xen_9pdev->rings[i].ring_order));
	336	}
	337	if (xen_9pdev->rings[i].intf != NULL) {
	338	xengnttab_unmap(xen_9pdev->xendev.gnttabdev,
	339	xen_9pdev->rings[i].intf,
	340	1);
	341	}
f23ef34a SS	342	if (xen_9pdev->rings[i].bh != NULL) {
	343	qemu_bh_delete(xen_9pdev->rings[i].bh);
	344	}
	345	}
e08d1e11 SS	346
	347	g_free(xen_9pdev->id);
	348	g_free(xen_9pdev->tag);
	349	g_free(xen_9pdev->path);
	350	g_free(xen_9pdev->security_model);
f23ef34a SS	351	g_free(xen_9pdev->rings);
f23ef34a SS	352	return 0;
b37eeb02 SS	353	}
	354
	355	static int xen_9pfs_connect(struct XenDevice *xendev)
	356	{
f23ef34a SS	357	int i;
	358	Xen9pfsDev *xen_9pdev = container_of(xendev, Xen9pfsDev, xendev);
	359	V9fsState *s = &xen_9pdev->state;
	360	QemuOpts *fsdev;
	361
	362	if (xenstore_read_fe_int(&xen_9pdev->xendev, "num-rings",
	363	&xen_9pdev->num_rings) == -1 \|\|
	364	xen_9pdev->num_rings > MAX_RINGS \|\| xen_9pdev->num_rings < 1) {
	365	return -1;
	366	}
	367
	368	xen_9pdev->rings = g_malloc0(xen_9pdev->num_rings * sizeof(Xen9pfsRing));
	369	for (i = 0; i < xen_9pdev->num_rings; i++) {
	370	char *str;
	371	int ring_order;
	372
	373	xen_9pdev->rings[i].priv = xen_9pdev;
	374	xen_9pdev->rings[i].evtchn = -1;
	375	xen_9pdev->rings[i].local_port = -1;
	376
	377	str = g_strdup_printf("ring-ref%u", i);
	378	if (xenstore_read_fe_int(&xen_9pdev->xendev, str,
	379	&xen_9pdev->rings[i].ref) == -1) {
c0c24b95	380	g_free(str);
f23ef34a SS	381	goto out;
	382	}
	383	g_free(str);
	384	str = g_strdup_printf("event-channel-%u", i);
	385	if (xenstore_read_fe_int(&xen_9pdev->xendev, str,
	386	&xen_9pdev->rings[i].evtchn) == -1) {
c0c24b95	387	g_free(str);
f23ef34a SS	388	goto out;
	389	}
	390	g_free(str);
	391
	392	xen_9pdev->rings[i].intf = xengnttab_map_grant_ref(
	393	xen_9pdev->xendev.gnttabdev,
	394	xen_9pdev->xendev.dom,
	395	xen_9pdev->rings[i].ref,
	396	PROT_READ \| PROT_WRITE);
	397	if (!xen_9pdev->rings[i].intf) {
	398	goto out;
	399	}
	400	ring_order = xen_9pdev->rings[i].intf->ring_order;
	401	if (ring_order > MAX_RING_ORDER) {
	402	goto out;
	403	}
	404	xen_9pdev->rings[i].ring_order = ring_order;
	405	xen_9pdev->rings[i].data = xengnttab_map_domain_grant_refs(
	406	xen_9pdev->xendev.gnttabdev,
	407	(1 << ring_order),
	408	xen_9pdev->xendev.dom,
	409	xen_9pdev->rings[i].intf->ref,
	410	PROT_READ \| PROT_WRITE);
	411	if (!xen_9pdev->rings[i].data) {
	412	goto out;
	413	}
	414	xen_9pdev->rings[i].ring.in = xen_9pdev->rings[i].data;
	415	xen_9pdev->rings[i].ring.out = xen_9pdev->rings[i].data +
	416	XEN_FLEX_RING_SIZE(ring_order);
	417
	418	xen_9pdev->rings[i].bh = qemu_bh_new(xen_9pfs_bh, &xen_9pdev->rings[i]);
	419	xen_9pdev->rings[i].out_cons = 0;
	420	xen_9pdev->rings[i].out_size = 0;
	421	xen_9pdev->rings[i].inprogress = false;
	422
	423
	424	xen_9pdev->rings[i].evtchndev = xenevtchn_open(NULL, 0);
	425	if (xen_9pdev->rings[i].evtchndev == NULL) {
	426	goto out;
	427	}
01cd90b6	428	qemu_set_cloexec(xenevtchn_fd(xen_9pdev->rings[i].evtchndev));
f23ef34a SS	429	xen_9pdev->rings[i].local_port = xenevtchn_bind_interdomain
	430	(xen_9pdev->rings[i].evtchndev,
	431	xendev->dom,
	432	xen_9pdev->rings[i].evtchn);
	433	if (xen_9pdev->rings[i].local_port == -1) {
	434	xen_pv_printf(xendev, 0,
	435	"xenevtchn_bind_interdomain failed port=%d\n",
	436	xen_9pdev->rings[i].evtchn);
	437	goto out;
	438	}
	439	xen_pv_printf(xendev, 2, "bind evtchn port %d\n", xendev->local_port);
	440	qemu_set_fd_handler(xenevtchn_fd(xen_9pdev->rings[i].evtchndev),
	441	xen_9pfs_evtchn_event, NULL, &xen_9pdev->rings[i]);
	442	}
	443
	444	xen_9pdev->security_model = xenstore_read_be_str(xendev, "security_model");
	445	xen_9pdev->path = xenstore_read_be_str(xendev, "path");
	446	xen_9pdev->id = s->fsconf.fsdev_id =
	447	g_strdup_printf("xen9p%d", xendev->dev);
	448	xen_9pdev->tag = s->fsconf.tag = xenstore_read_fe_str(xendev, "tag");
	449	v9fs_register_transport(s, &xen_9p_transport);
	450	fsdev = qemu_opts_create(qemu_find_opts("fsdev"),
	451	s->fsconf.tag,
	452	1, NULL);
	453	qemu_opt_set(fsdev, "fsdriver", "local", NULL);
	454	qemu_opt_set(fsdev, "path", xen_9pdev->path, NULL);
	455	qemu_opt_set(fsdev, "security_model", xen_9pdev->security_model, NULL);
	456	qemu_opts_set_id(fsdev, s->fsconf.fsdev_id);
	457	qemu_fsdev_add(fsdev);
	458	v9fs_device_realize_common(s, NULL);
	459
b37eeb02	460	return 0;
f23ef34a SS	461
	462	out:
	463	xen_9pfs_free(xendev);
	464	return -1;
b37eeb02 SS	465	}
	466
	467	static void xen_9pfs_alloc(struct XenDevice *xendev)
	468	{
f23ef34a SS	469	xenstore_write_be_str(xendev, "versions", VERSIONS);
	470	xenstore_write_be_int(xendev, "max-rings", MAX_RINGS);
	471	xenstore_write_be_int(xendev, "max-ring-page-order", MAX_RING_ORDER);
b37eeb02 SS	472	}
b37eeb02 SS	473
b37eeb02 SS	474	struct XenDevOps xen_9pfs_ops = {
	475	.size = sizeof(Xen9pfsDev),
	476	.flags = DEVOPS_FLAG_NEED_GNTDEV,
	477	.alloc = xen_9pfs_alloc,
	478	.init = xen_9pfs_init,
	479	.initialise = xen_9pfs_connect,
	480	.disconnect = xen_9pfs_disconnect,
	481	.free = xen_9pfs_free,
	482	};