Re: [PATCH v16 3/7] net/vmnet: implement shared mode (vmnet-shared)

[Akihiko Odaki <akihiko.odaki@gmail.com>]

On 2022/03/15 7:34, Akihiko Odaki wrote:
> On 2022/03/15 6:50, Vladislav Yaroshchuk wrote:
>> Thank you, Akihiko
>>
>> On Mon, Mar 14, 2022 at 10:46 PM Akihiko Odaki 
>> <akihiko.odaki@gmail.com <mailto:akihiko.odaki@gmail.com>> wrote:
>>
>>     On 2022/03/15 4:15, Vladislav Yaroshchuk wrote:
>>      > vmnet.framework supports iov, but writing more than
>>      > one iov into vmnet interface fails with
>>      > 'VMNET_INVALID_ARGUMENT'. Collecting provided iovs into
>>      > one and passing it to vmnet works fine. That's the
>>      > reason why receive_iov() left unimplemented. But it still
>>      > works with good enough performance having .receive()
>>      > implemented only.
>>      >
>>      > Signed-off-by: Phillip Tennen <phillip@axleos.com
>>     <mailto:phillip@axleos.com>>
>>      > Signed-off-by: Vladislav Yaroshchuk
>>     <Vladislav.Yaroshchuk@jetbrains.com
>>     <mailto:Vladislav.Yaroshchuk@jetbrains.com>>
>>      > ---
>>      >   net/vmnet-common.m | 298
>>     +++++++++++++++++++++++++++++++++++++++++++++
>>      >   net/vmnet-shared.c |  95 ++++++++++++++-
>>      >   net/vmnet_int.h    |  41 ++++++-
>>      >   3 files changed, 429 insertions(+), 5 deletions(-)
>>      >
>>      > diff --git a/net/vmnet-common.m b/net/vmnet-common.m
>>      > index 56612c72ce..20a33d2591 100644
>>      > --- a/net/vmnet-common.m
>>      > +++ b/net/vmnet-common.m
>>      > @@ -10,6 +10,8 @@
>>      >    */
>>      >
>>      >   #include "qemu/osdep.h"
>>      > +#include "qemu/main-loop.h"
>>      > +#include "qemu/log.h"
>>      >   #include "qapi/qapi-types-net.h"
>>      >   #include "vmnet_int.h"
>>      >   #include "clients.h"
>>      > @@ -17,4 +19,300 @@
>>      >   #include "qapi/error.h"
>>      >
>>      >   #include <vmnet/vmnet.h>
>>      > +#include <dispatch/dispatch.h>
>>      >
>>      > +static bool vmnet_qemu_send_wrapper(VmnetCommonState *s);
>>
>>     The names of vmnet_qemu_send_wrapper and vmnet_send_bh does not tell
>>     them apart well. Since only vmnet_send_bh does reading, its name may
>>     include "read" to clarify that. "wrapper" in vmnet_qemu_send_wrapper
>>     may
>>     be also misleading as it does more than just calling the underlying
>>     QEMU
>>     facility, but it also updates VmnetCommonState.
>>
>>
>> Ok, I'll think about how to name them better.
>>
>>      > +
>>      > +
>>      > +static void vmnet_send_completed(NetClientState *nc, ssize_t len)
>>      > +{
>>      > +    VmnetCommonState *s = DO_UPCAST(VmnetCommonState, nc, nc);
>>      > +    /* Complete sending packets left in VmnetCommonState 
>> buffers */
>>      > +    s->send_enabled = vmnet_qemu_send_wrapper(s);
>>
>>     It must qemu_bh_schedule(s->send_bh) after vmnet_qemu_send_wrapper.
>>
>>
>> Agree with you, thanks.
>>
>>     Also, send_enabled flag can be removed as explained in:
>>     https://www.mail-archive.com/qemu-devel@nongnu.org/msg873923.html
>>     <https://www.mail-archive.com/qemu-devel@nongnu.org/msg873923.html>
>>
>>
>> Not sure about this. Values of packets_send_current_pos
>> and packets_send_end_pos may be equal, but QEMU may be
>> not ready to receive new packets - the explanation:
>> 1. We are sending packets to QEMU with qemu_send_packet_async:
>>      packets_send_current_pos = 0
>>      packets_send_end_pos = 5
>> 2. All five packets (0, 1, 2, 3, 4) have been successfully sent to QEMU,
>>      but qemu_send_packet_async returned 0 "no more packets" after
>>      the last invocation
>> 3. In spite of this, all five packets are sent and
>>      packets_send_current_pos == packets_send_end_pos == 5
>> 4. It seems that "pointers are equal ->  QEMU is ready", but actually
>>      it is not.
>>
>> Also, hiding QEMU "ready"/"not ready" state behind pointers is a
>> bad choice I think. Having a concrete flag makes this more clear.
>> It provides understandability, not complexity (imho).
> 
> packets_send_current_pos must not be incremented if 
> qemu_send_packet_async returned 0. It must tell the position of the 
> packet currently being sent.

(And of course increment it in vmnet_send_completed instead.)

> 
> It would not hide the state, but it would rather make it clear that the 
> condition vmnet_send_bh can execute. If you see the current 
> implementation of vmnet_send_bh, you'll find the send_enabled flag, but 
> it does not tell the exact condition it requires to be enabled. You have 
> to then jump to all assignments for the flag to know it becomes true iff 
> every packets in the buffer are sent. It is obvious if vmnet_send_bh 
> directly states `if (packets_send_current_pos < packets_send_end_pos)`.
> 
> Eliminating the flag would also remove the possiblity of forgetting to 
> maintain the separate state.
> 
>>
>>       > send_enabled can be eliminated. When it is enabled, 
>> packets_send_pos
>>       > and packets_batch_size must be equal. They must not be equal
>>       > otherwise. packets_send_pos must represent the position of the
>>     packet
>>       > which is not sent yet, possibly in the process of sending.
>>       > vmnet_send_completed must call qemu_send_wrapper before 
>> scheduling
>>       > send_bh. bh_send should do nothing if s->packets_send_pos <
>>       > s->packets_batch_size.
>>
>>      > +}
>>      > +
>>      > +
>>      > +static bool vmnet_qemu_send_wrapper(VmnetCommonState *s) {
>>      > +    ssize_t size;
>>      > +
>>      > +    /*
>>      > +     * Packets to send lay in [current_pos..end_pos)
>>      > +     * (including current_pos, excluding end_pos)
>>      > +     */
>>      > +    while (s->packets_send_current_pos < 
>> s->packets_send_end_pos) {
>>      > +        size = qemu_send_packet_async(&s->nc,
>>      > +     s->iov_buf[s->packets_send_current_pos].iov_base,
>>      > +     s->packets_buf[s->packets_send_current_pos].vm_pkt_size,
>>      > +                                      vmnet_send_completed);
>>      > +        ++s->packets_send_current_pos;
>>      > +        if (size == 0) {
>>      > +            /* QEMU is not ready - wait for completion callback
>>     call */
>>      > +            return false;
>>      > +        }
>>      > +    }
>>      > +    return true;
>>      > +}
>>      > +
>>      > +
>>      > +static void vmnet_send_bh(void *opaque)
>>      > +{
>>      > +    NetClientState *nc = (NetClientState *) opaque;
>>      > +    VmnetCommonState *s = DO_UPCAST(VmnetCommonState, nc, nc);
>>      > +    struct vmpktdesc *packets = s->packets_buf;
>>      > +    vmnet_return_t status;
>>      > +    int i;
>>      > +
>>      > +    /*
>>      > +     * Do nothing if QEMU is not ready - wait
>>      > +     * for completion callback invocation
>>      > +     */
>>      > +    if (!s->send_enabled) {
>>      > +        return;
>>      > +    }
>>      > +
>>      > +    /* Read as many packets as present */
>>      > +    s->packets_send_current_pos = 0;
>>      > +    s->packets_send_end_pos = VMNET_PACKETS_LIMIT;
>>      > +    for (i = 0; i < s->packets_send_end_pos; ++i) {
>>      > +        packets[i].vm_pkt_size = s->max_packet_size;
>>      > +        packets[i].vm_pkt_iovcnt = 1;
>>      > +        packets[i].vm_flags = 0;
>>      > +    }
>>      > +
>>      > +    status = vmnet_read(s->vmnet_if, packets,
>>     &s->packets_send_end_pos);
>>      > +    if (status != VMNET_SUCCESS) {
>>      > +        error_printf("vmnet: read failed: %s\n",
>>      > +                     vmnet_status_map_str(status));
>>      > +        s->packets_send_current_pos = 0;
>>      > +        s->packets_send_end_pos = 0;
>>      > +        return;
>>      > +    }
>>      > +
>>      > +    /* Send packets to QEMU */
>>      > +    s->send_enabled = vmnet_qemu_send_wrapper(s);
>>      > +}
>>      > +
>>      > +
>>      > +static void vmnet_bufs_init(VmnetCommonState *s)
>>      > +{
>>      > +    struct vmpktdesc *packets = s->packets_buf;
>>      > +    struct iovec *iov = s->iov_buf;
>>      > +    int i;
>>      > +
>>      > +    for (i = 0; i < VMNET_PACKETS_LIMIT; ++i) {
>>      > +        iov[i].iov_len = s->max_packet_size;
>>      > +        iov[i].iov_base = g_malloc0(iov[i].iov_len);
>>      > +        packets[i].vm_pkt_iov = iov + i;
>>      > +    }
>>      > +}
>>      > +
>>      > +
>>      > +const char *vmnet_status_map_str(vmnet_return_t status)
>>      > +{
>>      > +    switch (status) {
>>      > +    case VMNET_SUCCESS:
>>      > +        return "success";
>>      > +    case VMNET_FAILURE:
>>      > +        return "general failure (possibly not enough 
>> privileges)";
>>      > +    case VMNET_MEM_FAILURE:
>>      > +        return "memory allocation failure";
>>      > +    case VMNET_INVALID_ARGUMENT:
>>      > +        return "invalid argument specified";
>>      > +    case VMNET_SETUP_INCOMPLETE:
>>      > +        return "interface setup is not complete";
>>      > +    case VMNET_INVALID_ACCESS:
>>      > +        return "invalid access, permission denied";
>>      > +    case VMNET_PACKET_TOO_BIG:
>>      > +        return "packet size is larger than MTU";
>>      > +    case VMNET_BUFFER_EXHAUSTED:
>>      > +        return "buffers exhausted in kernel";
>>      > +    case VMNET_TOO_MANY_PACKETS:
>>      > +        return "packet count exceeds limit";
>>      > +#if defined(MAC_OS_VERSION_11_0) && \
>>      > +    MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_VERSION_11_0
>>      > +    case VMNET_SHARING_SERVICE_BUSY:
>>      > +        return "conflict, sharing service is in use";
>>      > +#endif
>>      > +    default:
>>      > +        return "unknown vmnet error";
>>      > +    }
>>      > +}
>>      > +
>>      > +
>>      > +int vmnet_if_create(NetClientState *nc,
>>      > +                    xpc_object_t if_desc,
>>      > +                    Error **errp)
>>      > +{
>>      > +    VmnetCommonState *s = DO_UPCAST(VmnetCommonState, nc, nc);
>>      > +    dispatch_semaphore_t if_created_sem =
>>     dispatch_semaphore_create(0);
>>      > +    __block vmnet_return_t if_status;
>>      > +
>>      > +    s->if_queue = dispatch_queue_create(
>>      > +        "org.qemu.vmnet.if_queue",
>>      > +        DISPATCH_QUEUE_SERIAL
>>      > +    );
>>      > +
>>      > +    xpc_dictionary_set_bool(
>>      > +        if_desc,
>>      > +        vmnet_allocate_mac_address_key,
>>      > +        false
>>      > +    );
>>      > +
>>      > +#ifdef DEBUG
>>      > +    qemu_log("vmnet.start.interface_desc:\n");
>>      > +    xpc_dictionary_apply(if_desc,
>>      > +                         ^bool(const char *k, xpc_object_t v) {
>>      > +                             char *desc = 
>> xpc_copy_description(v);
>>      > +                             qemu_log("  %s=%s\n", k, desc);
>>      > +                             free(desc);
>>      > +                             return true;
>>      > +                         });
>>      > +#endif /* DEBUG */
>>      > +
>>      > +    s->vmnet_if = vmnet_start_interface(
>>      > +        if_desc,
>>      > +        s->if_queue,
>>      > +        ^(vmnet_return_t status, xpc_object_t interface_param) {
>>      > +            if_status = status;
>>      > +            if (status != VMNET_SUCCESS || !interface_param) {
>>      > +                dispatch_semaphore_signal(if_created_sem);
>>      > +                return;
>>      > +            }
>>      > +
>>      > +#ifdef DEBUG
>>      > +            qemu_log("vmnet.start.interface_param:\n");
>>      > +            xpc_dictionary_apply(interface_param,
>>      > +                                 ^bool(const char *k,
>>     xpc_object_t v) {
>>      > +                                     char *desc =
>>     xpc_copy_description(v);
>>      > +                                     qemu_log("  %s=%s\n", k, 
>> desc);
>>      > +                                     free(desc);
>>      > +                                     return true;
>>      > +                                 });
>>      > +#endif /* DEBUG */
>>      > +
>>      > +            s->mtu = xpc_dictionary_get_uint64(
>>      > +                interface_param,
>>      > +                vmnet_mtu_key);
>>      > +            s->max_packet_size = xpc_dictionary_get_uint64(
>>      > +                interface_param,
>>      > +                vmnet_max_packet_size_key);
>>      > +
>>      > +            dispatch_semaphore_signal(if_created_sem);
>>      > +        });
>>      > +
>>      > +    if (s->vmnet_if == NULL) {
>>      > +        dispatch_release(s->if_queue);
>>      > +        dispatch_release(if_created_sem);
>>      > +        error_setg(errp,
>>      > +                   "unable to create interface with requested
>>     params");
>>      > +        return -1;
>>      > +    }
>>      > +
>>      > +    dispatch_semaphore_wait(if_created_sem, 
>> DISPATCH_TIME_FOREVER);
>>      > +    dispatch_release(if_created_sem);
>>      > +
>>      > +    if (if_status != VMNET_SUCCESS) {
>>      > +        dispatch_release(s->if_queue);
>>      > +        error_setg(errp,
>>      > +                   "cannot create vmnet interface: %s",
>>      > +                   vmnet_status_map_str(if_status));
>>      > +        return -1;
>>      > +    }
>>      > +
>>      > +    s->send_bh = aio_bh_new(qemu_get_aio_context(),
>>     vmnet_send_bh, nc);
>>      > +    s->send_enabled = true;
>>      > +    vmnet_bufs_init(s);
>>      > +
>>      > +    vmnet_interface_set_event_callback(
>>      > +        s->vmnet_if,
>>      > +        VMNET_INTERFACE_PACKETS_AVAILABLE,
>>      > +        s->if_queue,
>>      > +        ^(interface_event_t event_id, xpc_object_t event) {
>>      > +            assert(event_id == 
>> VMNET_INTERFACE_PACKETS_AVAILABLE);
>>      > +            /*
>>      > +             * This function is being called from a non qemu
>>     thread, so
>>      > +             * we only schedule a BH, and do the rest of the io
>>     completion
>>      > +             * handling from vmnet_send_bh() which runs in a
>>     qemu context.
>>      > +             */
>>      > +            qemu_bh_schedule(s->send_bh);
>>      > +        });
>>      > +
>>      > +    return 0;
>>      > +}
>>      > +
>>      > +
>>      > +ssize_t vmnet_receive_common(NetClientState *nc,
>>      > +                             const uint8_t *buf,
>>      > +                             size_t size)
>>      > +{
>>      > +    VmnetCommonState *s = DO_UPCAST(VmnetCommonState, nc, nc);
>>      > +    struct vmpktdesc packet;
>>      > +    struct iovec iov;
>>      > +    int pkt_cnt;
>>      > +    vmnet_return_t if_status;
>>      > +
>>      > +    if (size > s->max_packet_size) {
>>      > +        warn_report("vmnet: packet is too big, %zu > %" PRIu64,
>>      > +        packet.vm_pkt_size,
>>      > +        s->max_packet_size);
>>      > +        return -1;
>>      > +    }
>>      > +
>>      > +    iov.iov_base = (char *) buf;
>>      > +    iov.iov_len = size;
>>      > +
>>      > +    packet.vm_pkt_iovcnt = 1;
>>      > +    packet.vm_flags = 0;
>>      > +    packet.vm_pkt_size = size;
>>      > +    packet.vm_pkt_iov = &iov;
>>      > +    pkt_cnt = 1;
>>      > +
>>      > +    if_status = vmnet_write(s->vmnet_if, &packet, &pkt_cnt);
>>      > +    if (if_status != VMNET_SUCCESS) {
>>      > +        error_report("vmnet: write error: %s\n",
>>      > +                     vmnet_status_map_str(if_status));
>>      > +        return -1;
>>      > +    }
>>      > +
>>      > +    if (if_status == VMNET_SUCCESS && pkt_cnt) {
>>
>>     `if_status == VMNET_SUCCESS` is redundant.
>>
>>
>> Missed this, thanks.
>>
>>     Regards,
>>     Akihiko Odaki
>>
>>      > +        return size;
>>      > +    }
>>      > +    return 0;
>>      > +}
>>      > +
>>      > +
>>      > +void vmnet_cleanup_common(NetClientState *nc)
>>      > +{
>>      > +    VmnetCommonState *s = DO_UPCAST(VmnetCommonState, nc, nc);
>>      > +    dispatch_semaphore_t if_stopped_sem;
>>      > +
>>      > +    if (s->vmnet_if == NULL) {
>>      > +        return;
>>      > +    }
>>      > +
>>      > +    vmnet_interface_set_event_callback(
>>      > +        s->vmnet_if,
>>      > +        VMNET_INTERFACE_PACKETS_AVAILABLE,
>>      > +        NULL,
>>      > +        NULL);
>>
>>     I don't think this vmnet_interface_set_event_callback call is 
>> necessary.
>>
>>
>> I kept in mind that vmnet processing lives in a separate thread
>> and while cleanup it may continue receiving packets. While the
>> queue is not empty, vmnet_stop_interface hangs. Unregistering
>> callback ensures that this queue will be emptied asap.
>>
>> It will work without vmnet_interface_set_event_callback here,
>> but I think it's better to be respectful to vmnet and clean everything
>> we can :)
> 
> You may put qemu_purge_queued_packets after vmnet_stop_interface if you 
> think about the case it keeps receving packets while cleaning up since 
> it is the only thing it does before calling vmnet_stop_interface. 
> vmnet_stop_interface would then stop things in the proper order. It may 
> decide to stop event callbacks first. Otherwise, it may decide to stop 
> some internal heavy functionality first. It is up to vmnet.framework.
> 
> Regards,
> Akihiko Odaki
> 
>> Thank you!
>>
>> Best Regards,
>>
>> Vladislav
>>
>>      > +
>>      > +    qemu_purge_queued_packets(nc);
>>      > +
>>      > +    if_stopped_sem = dispatch_semaphore_create(0);
>>      > +    vmnet_stop_interface(
>>      > +        s->vmnet_if,
>>      > +        s->if_queue,
>>      > +        ^(vmnet_return_t status) {
>>      > +            assert(status == VMNET_SUCCESS);
>>      > +            dispatch_semaphore_signal(if_stopped_sem);
>>      > +        });
>>      > +    dispatch_semaphore_wait(if_stopped_sem, 
>> DISPATCH_TIME_FOREVER);
>>      > +
>>      > +    qemu_bh_delete(s->send_bh);
>>      > +    dispatch_release(if_stopped_sem);
>>      > +    dispatch_release(s->if_queue);
>>      > +
>>      > +    for (int i = 0; i < VMNET_PACKETS_LIMIT; ++i) {
>>      > +        g_free(s->iov_buf[i].iov_base);
>>      > +    }
>>      > +}
>>      > diff --git a/net/vmnet-shared.c b/net/vmnet-shared.c
>>      > index f07afaaf21..2f4eb1db2d 100644
>>      > --- a/net/vmnet-shared.c
>>      > +++ b/net/vmnet-shared.c
>>      > @@ -10,16 +10,103 @@
>>      >
>>      >   #include "qemu/osdep.h"
>>      >   #include "qapi/qapi-types-net.h"
>>      > +#include "qapi/error.h"
>>      >   #include "vmnet_int.h"
>>      >   #include "clients.h"
>>      > -#include "qemu/error-report.h"
>>      > -#include "qapi/error.h"
>>      >
>>      >   #include <vmnet/vmnet.h>
>>      >
>>      > +typedef struct VmnetSharedState {
>>      > +    VmnetCommonState cs;
>>      > +} VmnetSharedState;
>>      > +
>>      > +
>>      > +static bool validate_options(const Netdev *netdev, Error **errp)
>>      > +{
>>      > +    const NetdevVmnetSharedOptions *options =
>>     &(netdev->u.vmnet_shared);
>>      > +
>>      > +#if !defined(MAC_OS_VERSION_11_0) || \
>>      > +    MAC_OS_X_VERSION_MIN_REQUIRED < MAC_OS_VERSION_11_0
>>      > +    if (options->has_isolated) {
>>      > +        error_setg(errp,
>>      > +                   "vmnet-shared.isolated feature is "
>>      > +                   "unavailable: outdated vmnet.framework API");
>>      > +        return false;
>>      > +    }
>>      > +#endif
>>      > +
>>      > +    if ((options->has_start_address ||
>>      > +         options->has_end_address ||
>>      > +         options->has_subnet_mask) &&
>>      > +        !(options->has_start_address &&
>>      > +          options->has_end_address &&
>>      > +          options->has_subnet_mask)) {
>>      > +        error_setg(errp,
>>      > +                   "'start-address', 'end-address', 
>> 'subnet-mask' "
>>      > +                   "should be provided together"
>>      > +        );
>>      > +        return false;
>>      > +    }
>>      > +
>>      > +    return true;
>>      > +}
>>      > +
>>      > +static xpc_object_t build_if_desc(const Netdev *netdev)
>>      > +{
>>      > +    const NetdevVmnetSharedOptions *options =
>>     &(netdev->u.vmnet_shared);
>>      > +    xpc_object_t if_desc = xpc_dictionary_create(NULL, NULL, 0);
>>      > +
>>      > +    xpc_dictionary_set_uint64(
>>      > +        if_desc,
>>      > +        vmnet_operation_mode_key,
>>      > +        VMNET_SHARED_MODE
>>      > +    );
>>      > +
>>      > +    if (options->has_nat66_prefix) {
>>      > +        xpc_dictionary_set_string(if_desc,
>>      > +                                  vmnet_nat66_prefix_key,
>>      > +                                  options->nat66_prefix);
>>      > +    }
>>      > +
>>      > +    if (options->has_start_address) {
>>      > +        xpc_dictionary_set_string(if_desc,
>>      > +                                  vmnet_start_address_key,
>>      > +                                  options->start_address);
>>      > +        xpc_dictionary_set_string(if_desc,
>>      > +                                  vmnet_end_address_key,
>>      > +                                  options->end_address);
>>      > +        xpc_dictionary_set_string(if_desc,
>>      > +                                  vmnet_subnet_mask_key,
>>      > +                                  options->subnet_mask);
>>      > +    }
>>      > +
>>      > +#if defined(MAC_OS_VERSION_11_0) && \
>>      > +    MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_VERSION_11_0
>>      > +    xpc_dictionary_set_bool(
>>      > +        if_desc,
>>      > +        vmnet_enable_isolation_key,
>>      > +        options->isolated
>>      > +    );
>>      > +#endif
>>      > +
>>      > +    return if_desc;
>>      > +}
>>      > +
>>      > +static NetClientInfo net_vmnet_shared_info = {
>>      > +    .type = NET_CLIENT_DRIVER_VMNET_SHARED,
>>      > +    .size = sizeof(VmnetSharedState),
>>      > +    .receive = vmnet_receive_common,
>>      > +    .cleanup = vmnet_cleanup_common,
>>      > +};
>>      > +
>>      >   int net_init_vmnet_shared(const Netdev *netdev, const char 
>> *name,
>>      >                             NetClientState *peer, Error **errp)
>>      >   {
>>      > -  error_setg(errp, "vmnet-shared is not implemented yet");
>>      > -  return -1;
>>      > +    NetClientState *nc = 
>> qemu_new_net_client(&net_vmnet_shared_info,
>>      > +                                             peer,
>>     "vmnet-shared", name);
>>      > +    if (!validate_options(netdev, errp)) {
>>      > +        g_assert_not_reached();
>>      > +        return -1;
>>      > +    }
>>      > +    return vmnet_if_create(nc, build_if_desc(netdev), errp);
>>      >   }
>>      > diff --git a/net/vmnet_int.h b/net/vmnet_int.h
>>      > index aac4d5af64..8f3321ef3e 100644
>>      > --- a/net/vmnet_int.h
>>      > +++ b/net/vmnet_int.h
>>      > @@ -15,11 +15,50 @@
>>      >   #include "clients.h"
>>      >
>>      >   #include <vmnet/vmnet.h>
>>      > +#include <dispatch/dispatch.h>
>>      > +
>>      > +/**
>>      > + *  From vmnet.framework documentation
>>      > + *
>>      > + *  Each read/write call allows up to 200 packets to be
>>      > + *  read or written for a maximum of 256KB.
>>      > + *
>>      > + *  Each packet written should be a complete
>>      > + *  ethernet frame.
>>      > + *
>>      > + * https://developer.apple.com/documentation/vmnet
>>     <https://developer.apple.com/documentation/vmnet>
>>      > + */
>>      > +#define VMNET_PACKETS_LIMIT 200
>>      >
>>      >   typedef struct VmnetCommonState {
>>      > -  NetClientState nc;
>>      > +    NetClientState nc;
>>      > +    interface_ref vmnet_if;
>>      > +
>>      > +    uint64_t mtu;
>>      > +    uint64_t max_packet_size;
>>      >
>>      > +    dispatch_queue_t if_queue;
>>      > +
>>      > +    QEMUBH *send_bh;
>>      > +    bool send_enabled;
>>      > +
>>      > +    struct vmpktdesc packets_buf[VMNET_PACKETS_LIMIT];
>>      > +    int packets_send_current_pos;
>>      > +    int packets_send_end_pos;
>>      > +
>>      > +    struct iovec iov_buf[VMNET_PACKETS_LIMIT];
>>      >   } VmnetCommonState;
>>      >
>>      > +const char *vmnet_status_map_str(vmnet_return_t status);
>>      > +
>>      > +int vmnet_if_create(NetClientState *nc,
>>      > +                    xpc_object_t if_desc,
>>      > +                    Error **errp);
>>      > +
>>      > +ssize_t vmnet_receive_common(NetClientState *nc,
>>      > +                             const uint8_t *buf,
>>      > +                             size_t size);
>>      > +
>>      > +void vmnet_cleanup_common(NetClientState *nc);
>>      >
>>      >   #endif /* VMNET_INT_H */
>>
>