1 // SPDX-License-Identifier: GPL-2.0-only
3 * Hyper-V transport for vsock
5 * Hyper-V Sockets supplies a byte-stream based communication mechanism
6 * between the host and the VM. This driver implements the necessary
7 * support in the VM by introducing the new vsock transport.
9 * Copyright (c) 2017, Microsoft Corporation.
11 #include <linux/module.h>
12 #include <linux/vmalloc.h>
13 #include <linux/hyperv.h>
15 #include <net/af_vsock.h>
16 #include <asm/hyperv-tlfs.h>
18 /* Older (VMBUS version 'VERSION_WIN10' or before) Windows hosts have some
19 * stricter requirements on the hv_sock ring buffer size of six 4K pages.
20 * hyperv-tlfs defines HV_HYP_PAGE_SIZE as 4K. Newer hosts don't have this
21 * limitation; but, keep the defaults the same for compat.
23 #define RINGBUFFER_HVS_RCV_SIZE (HV_HYP_PAGE_SIZE * 6)
24 #define RINGBUFFER_HVS_SND_SIZE (HV_HYP_PAGE_SIZE * 6)
25 #define RINGBUFFER_HVS_MAX_SIZE (HV_HYP_PAGE_SIZE * 64)
27 /* The MTU is 16KB per the host side's design */
28 #define HVS_MTU_SIZE (1024 * 16)
30 /* How long to wait for graceful shutdown of a connection */
31 #define HVS_CLOSE_TIMEOUT (8 * HZ)
33 struct vmpipe_proto_header {
38 /* For recv, we use the VMBus in-place packet iterator APIs to directly copy
39 * data from the ringbuffer into the userspace buffer.
42 /* The header before the payload data */
43 struct vmpipe_proto_header hdr;
46 u8 data[HVS_MTU_SIZE];
49 /* We can send up to HVS_MTU_SIZE bytes of payload to the host, but let's use
50 * a smaller size, i.e. HVS_SEND_BUF_SIZE, to maximize concurrency between the
51 * guest and the host processing as one VMBUS packet is the smallest processing
54 * Note: the buffer can be eliminated in the future when we add new VMBus
55 * ringbuffer APIs that allow us to directly copy data from userspace buffer
56 * to VMBus ringbuffer.
58 #define HVS_SEND_BUF_SIZE \
59 (HV_HYP_PAGE_SIZE - sizeof(struct vmpipe_proto_header))
62 /* The header before the payload data */
63 struct vmpipe_proto_header hdr;
66 u8 data[HVS_SEND_BUF_SIZE];
69 #define HVS_HEADER_LEN (sizeof(struct vmpacket_descriptor) + \
70 sizeof(struct vmpipe_proto_header))
72 /* See 'prev_indices' in hv_ringbuffer_read(), hv_ringbuffer_write(), and
73 * __hv_pkt_iter_next().
75 #define VMBUS_PKT_TRAILER_SIZE (sizeof(u64))
77 #define HVS_PKT_LEN(payload_len) (HVS_HEADER_LEN + \
78 ALIGN((payload_len), 8) + \
79 VMBUS_PKT_TRAILER_SIZE)
81 union hvs_service_id {
85 unsigned int svm_port;
86 unsigned char b[sizeof(guid_t) - sizeof(unsigned int)];
90 /* Per-socket state (accessed via vsk->trans) */
92 struct vsock_sock *vsk;
97 struct vmbus_channel *chan;
98 struct vmpacket_descriptor *recv_desc;
100 /* The length of the payload not delivered to userland yet */
102 /* The offset of the payload */
105 /* Have we sent the zero-length packet (FIN)? */
109 /* In the VM, we support Hyper-V Sockets with AF_VSOCK, and the endpoint is
110 * <cid, port> (see struct sockaddr_vm). Note: cid is not really used here:
111 * when we write apps to connect to the host, we can only use VMADDR_CID_ANY
112 * or VMADDR_CID_HOST (both are equivalent) as the remote cid, and when we
113 * write apps to bind() & listen() in the VM, we can only use VMADDR_CID_ANY
116 * On the host, Hyper-V Sockets are supported by Winsock AF_HYPERV:
117 * https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/user-
118 * guide/make-integration-service, and the endpoint is <VmID, ServiceId> with
119 * the below sockaddr:
123 * ADDRESS_FAMILY Family;
128 * Note: VmID is not used by Linux VM and actually it isn't transmitted via
129 * VMBus, because here it's obvious the host and the VM can easily identify
130 * each other. Though the VmID is useful on the host, especially in the case
131 * of Windows container, Linux VM doesn't need it at all.
133 * To make use of the AF_VSOCK infrastructure in Linux VM, we have to limit
134 * the available GUID space of SOCKADDR_HV so that we can create a mapping
135 * between AF_VSOCK port and SOCKADDR_HV Service GUID. The rule of writing
136 * Hyper-V Sockets apps on the host and in Linux VM is:
138 ****************************************************************************
139 * The only valid Service GUIDs, from the perspectives of both the host and *
140 * Linux VM, that can be connected by the other end, must conform to this *
141 * format: <port>-facb-11e6-bd58-64006a7986d3, and the "port" must be in *
142 * this range [0, 0x7FFFFFFF]. *
143 ****************************************************************************
145 * When we write apps on the host to connect(), the GUID ServiceID is used.
146 * When we write apps in Linux VM to connect(), we only need to specify the
147 * port and the driver will form the GUID and use that to request the host.
149 * From the perspective of Linux VM:
150 * 1. the local ephemeral port (i.e. the local auto-bound port when we call
151 * connect() without explicit bind()) is generated by __vsock_bind_stream(),
152 * and the range is [1024, 0xFFFFFFFF).
153 * 2. the remote ephemeral port (i.e. the auto-generated remote port for
154 * a connect request initiated by the host's connect()) is generated by
155 * hvs_remote_addr_init() and the range is [0x80000000, 0xFFFFFFFF).
158 #define MAX_LISTEN_PORT ((u32)0x7FFFFFFF)
159 #define MAX_VM_LISTEN_PORT MAX_LISTEN_PORT
160 #define MAX_HOST_LISTEN_PORT MAX_LISTEN_PORT
161 #define MIN_HOST_EPHEMERAL_PORT (MAX_HOST_LISTEN_PORT + 1)
163 /* 00000000-facb-11e6-bd58-64006a7986d3 */
164 static const guid_t srv_id_template =
165 GUID_INIT(0x00000000, 0xfacb, 0x11e6, 0xbd, 0x58,
166 0x64, 0x00, 0x6a, 0x79, 0x86, 0xd3);
168 static bool is_valid_srv_id(const guid_t *id)
170 return !memcmp(&id->b[4], &srv_id_template.b[4], sizeof(guid_t) - 4);
173 static unsigned int get_port_by_srv_id(const guid_t *svr_id)
175 return *((unsigned int *)svr_id);
178 static void hvs_addr_init(struct sockaddr_vm *addr, const guid_t *svr_id)
180 unsigned int port = get_port_by_srv_id(svr_id);
182 vsock_addr_init(addr, VMADDR_CID_ANY, port);
185 static void hvs_remote_addr_init(struct sockaddr_vm *remote,
186 struct sockaddr_vm *local)
188 static u32 host_ephemeral_port = MIN_HOST_EPHEMERAL_PORT;
191 vsock_addr_init(remote, VMADDR_CID_ANY, VMADDR_PORT_ANY);
195 if (host_ephemeral_port < MIN_HOST_EPHEMERAL_PORT ||
196 host_ephemeral_port == VMADDR_PORT_ANY)
197 host_ephemeral_port = MIN_HOST_EPHEMERAL_PORT;
199 remote->svm_port = host_ephemeral_port++;
201 sk = vsock_find_connected_socket(remote, local);
203 /* Found an available ephemeral port */
207 /* Release refcnt got in vsock_find_connected_socket */
212 static void hvs_set_channel_pending_send_size(struct vmbus_channel *chan)
214 set_channel_pending_send_size(chan,
215 HVS_PKT_LEN(HVS_SEND_BUF_SIZE));
220 static bool hvs_channel_readable(struct vmbus_channel *chan)
222 u32 readable = hv_get_bytes_to_read(&chan->inbound);
224 /* 0-size payload means FIN */
225 return readable >= HVS_PKT_LEN(0);
228 static int hvs_channel_readable_payload(struct vmbus_channel *chan)
230 u32 readable = hv_get_bytes_to_read(&chan->inbound);
232 if (readable > HVS_PKT_LEN(0)) {
233 /* At least we have 1 byte to read. We don't need to return
234 * the exact readable bytes: see vsock_stream_recvmsg() ->
235 * vsock_stream_has_data().
240 if (readable == HVS_PKT_LEN(0)) {
241 /* 0-size payload means FIN */
245 /* No payload or FIN */
249 static size_t hvs_channel_writable_bytes(struct vmbus_channel *chan)
251 u32 writeable = hv_get_bytes_to_write(&chan->outbound);
254 /* The ringbuffer mustn't be 100% full, and we should reserve a
255 * zero-length-payload packet for the FIN: see hv_ringbuffer_write()
256 * and hvs_shutdown().
258 if (writeable <= HVS_PKT_LEN(1) + HVS_PKT_LEN(0))
261 ret = writeable - HVS_PKT_LEN(1) - HVS_PKT_LEN(0);
263 return round_down(ret, 8);
266 static int hvs_send_data(struct vmbus_channel *chan,
267 struct hvs_send_buf *send_buf, size_t to_write)
269 send_buf->hdr.pkt_type = 1;
270 send_buf->hdr.data_size = to_write;
271 return vmbus_sendpacket(chan, &send_buf->hdr,
272 sizeof(send_buf->hdr) + to_write,
273 0, VM_PKT_DATA_INBAND, 0);
276 static void hvs_channel_cb(void *ctx)
278 struct sock *sk = (struct sock *)ctx;
279 struct vsock_sock *vsk = vsock_sk(sk);
280 struct hvsock *hvs = vsk->trans;
281 struct vmbus_channel *chan = hvs->chan;
283 if (hvs_channel_readable(chan))
284 sk->sk_data_ready(sk);
286 if (hv_get_bytes_to_write(&chan->outbound) > 0)
287 sk->sk_write_space(sk);
290 static void hvs_do_close_lock_held(struct vsock_sock *vsk,
293 struct sock *sk = sk_vsock(vsk);
295 sock_set_flag(sk, SOCK_DONE);
296 vsk->peer_shutdown = SHUTDOWN_MASK;
297 if (vsock_stream_has_data(vsk) <= 0)
298 sk->sk_state = TCP_CLOSING;
299 sk->sk_state_change(sk);
300 if (vsk->close_work_scheduled &&
301 (!cancel_timeout || cancel_delayed_work(&vsk->close_work))) {
302 vsk->close_work_scheduled = false;
303 vsock_remove_sock(vsk);
305 /* Release the reference taken while scheduling the timeout */
310 static void hvs_close_connection(struct vmbus_channel *chan)
312 struct sock *sk = get_per_channel_state(chan);
315 hvs_do_close_lock_held(vsock_sk(sk), true);
318 /* Release the refcnt for the channel that's opened in
319 * hvs_open_connection().
324 static void hvs_open_connection(struct vmbus_channel *chan)
326 guid_t *if_instance, *if_type;
327 unsigned char conn_from_host;
329 struct sockaddr_vm addr;
330 struct sock *sk, *new = NULL;
331 struct vsock_sock *vnew = NULL;
332 struct hvsock *hvs = NULL;
333 struct hvsock *hvs_new = NULL;
338 if_type = &chan->offermsg.offer.if_type;
339 if_instance = &chan->offermsg.offer.if_instance;
340 conn_from_host = chan->offermsg.offer.u.pipe.user_def[0];
342 /* The host or the VM should only listen on a port in
343 * [0, MAX_LISTEN_PORT]
345 if (!is_valid_srv_id(if_type) ||
346 get_port_by_srv_id(if_type) > MAX_LISTEN_PORT)
349 hvs_addr_init(&addr, conn_from_host ? if_type : if_instance);
350 sk = vsock_find_bound_socket(&addr);
355 if ((conn_from_host && sk->sk_state != TCP_LISTEN) ||
356 (!conn_from_host && sk->sk_state != TCP_SYN_SENT))
359 if (conn_from_host) {
360 if (sk->sk_ack_backlog >= sk->sk_max_ack_backlog)
363 new = __vsock_create(sock_net(sk), NULL, sk, GFP_KERNEL,
368 new->sk_state = TCP_SYN_SENT;
369 vnew = vsock_sk(new);
370 hvs_new = vnew->trans;
371 hvs_new->chan = chan;
373 hvs = vsock_sk(sk)->trans;
377 set_channel_read_mode(chan, HV_CALL_DIRECT);
379 /* Use the socket buffer sizes as hints for the VMBUS ring size. For
380 * server side sockets, 'sk' is the parent socket and thus, this will
381 * allow the child sockets to inherit the size from the parent. Keep
382 * the mins to the default value and align to page size as per VMBUS
384 * For the max, the socket core library will limit the socket buffer
385 * size that can be set by the user, but, since currently, the hv_sock
386 * VMBUS ring buffer is physically contiguous allocation, restrict it
388 * Older versions of hv_sock host side code cannot handle bigger VMBUS
389 * ring buffer size. Use the version number to limit the change to newer
392 if (vmbus_proto_version < VERSION_WIN10_V5) {
393 sndbuf = RINGBUFFER_HVS_SND_SIZE;
394 rcvbuf = RINGBUFFER_HVS_RCV_SIZE;
396 sndbuf = max_t(int, sk->sk_sndbuf, RINGBUFFER_HVS_SND_SIZE);
397 sndbuf = min_t(int, sndbuf, RINGBUFFER_HVS_MAX_SIZE);
398 sndbuf = ALIGN(sndbuf, HV_HYP_PAGE_SIZE);
399 rcvbuf = max_t(int, sk->sk_rcvbuf, RINGBUFFER_HVS_RCV_SIZE);
400 rcvbuf = min_t(int, rcvbuf, RINGBUFFER_HVS_MAX_SIZE);
401 rcvbuf = ALIGN(rcvbuf, HV_HYP_PAGE_SIZE);
404 ret = vmbus_open(chan, sndbuf, rcvbuf, NULL, 0, hvs_channel_cb,
405 conn_from_host ? new : sk);
407 if (conn_from_host) {
408 hvs_new->chan = NULL;
416 set_per_channel_state(chan, conn_from_host ? new : sk);
418 /* This reference will be dropped by hvs_close_connection(). */
419 sock_hold(conn_from_host ? new : sk);
420 vmbus_set_chn_rescind_callback(chan, hvs_close_connection);
422 /* Set the pending send size to max packet size to always get
423 * notifications from the host when there is enough writable space.
424 * The host is optimized to send notifications only when the pending
425 * size boundary is crossed, and not always.
427 hvs_set_channel_pending_send_size(chan);
429 if (conn_from_host) {
430 new->sk_state = TCP_ESTABLISHED;
431 sk->sk_ack_backlog++;
433 hvs_addr_init(&vnew->local_addr, if_type);
434 hvs_remote_addr_init(&vnew->remote_addr, &vnew->local_addr);
436 hvs_new->vm_srv_id = *if_type;
437 hvs_new->host_srv_id = *if_instance;
439 vsock_insert_connected(vnew);
441 vsock_enqueue_accept(sk, new);
443 sk->sk_state = TCP_ESTABLISHED;
444 sk->sk_socket->state = SS_CONNECTED;
446 vsock_insert_connected(vsock_sk(sk));
449 sk->sk_state_change(sk);
452 /* Release refcnt obtained when we called vsock_find_bound_socket() */
458 static u32 hvs_get_local_cid(void)
460 return VMADDR_CID_ANY;
463 static int hvs_sock_init(struct vsock_sock *vsk, struct vsock_sock *psk)
466 struct sock *sk = sk_vsock(vsk);
468 hvs = kzalloc(sizeof(*hvs), GFP_KERNEL);
474 sk->sk_sndbuf = RINGBUFFER_HVS_SND_SIZE;
475 sk->sk_rcvbuf = RINGBUFFER_HVS_RCV_SIZE;
479 static int hvs_connect(struct vsock_sock *vsk)
481 union hvs_service_id vm, host;
482 struct hvsock *h = vsk->trans;
484 vm.srv_id = srv_id_template;
485 vm.svm_port = vsk->local_addr.svm_port;
486 h->vm_srv_id = vm.srv_id;
488 host.srv_id = srv_id_template;
489 host.svm_port = vsk->remote_addr.svm_port;
490 h->host_srv_id = host.srv_id;
492 return vmbus_send_tl_connect_request(&h->vm_srv_id, &h->host_srv_id);
495 static void hvs_shutdown_lock_held(struct hvsock *hvs, int mode)
497 struct vmpipe_proto_header hdr;
499 if (hvs->fin_sent || !hvs->chan)
502 /* It can't fail: see hvs_channel_writable_bytes(). */
503 (void)hvs_send_data(hvs->chan, (struct hvs_send_buf *)&hdr, 0);
504 hvs->fin_sent = true;
507 static int hvs_shutdown(struct vsock_sock *vsk, int mode)
509 struct sock *sk = sk_vsock(vsk);
511 if (!(mode & SEND_SHUTDOWN))
515 hvs_shutdown_lock_held(vsk->trans, mode);
520 static void hvs_close_timeout(struct work_struct *work)
522 struct vsock_sock *vsk =
523 container_of(work, struct vsock_sock, close_work.work);
524 struct sock *sk = sk_vsock(vsk);
528 if (!sock_flag(sk, SOCK_DONE))
529 hvs_do_close_lock_held(vsk, false);
531 vsk->close_work_scheduled = false;
536 /* Returns true, if it is safe to remove socket; false otherwise */
537 static bool hvs_close_lock_held(struct vsock_sock *vsk)
539 struct sock *sk = sk_vsock(vsk);
541 if (!(sk->sk_state == TCP_ESTABLISHED ||
542 sk->sk_state == TCP_CLOSING))
545 if ((sk->sk_shutdown & SHUTDOWN_MASK) != SHUTDOWN_MASK)
546 hvs_shutdown_lock_held(vsk->trans, SHUTDOWN_MASK);
548 if (sock_flag(sk, SOCK_DONE))
551 /* This reference will be dropped by the delayed close routine */
553 INIT_DELAYED_WORK(&vsk->close_work, hvs_close_timeout);
554 vsk->close_work_scheduled = true;
555 schedule_delayed_work(&vsk->close_work, HVS_CLOSE_TIMEOUT);
559 static void hvs_release(struct vsock_sock *vsk)
561 struct sock *sk = sk_vsock(vsk);
564 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
565 remove_sock = hvs_close_lock_held(vsk);
568 vsock_remove_sock(vsk);
571 static void hvs_destruct(struct vsock_sock *vsk)
573 struct hvsock *hvs = vsk->trans;
574 struct vmbus_channel *chan = hvs->chan;
577 vmbus_hvsock_device_unregister(chan);
582 static int hvs_dgram_bind(struct vsock_sock *vsk, struct sockaddr_vm *addr)
587 static int hvs_dgram_dequeue(struct vsock_sock *vsk, struct msghdr *msg,
588 size_t len, int flags)
593 static int hvs_dgram_enqueue(struct vsock_sock *vsk,
594 struct sockaddr_vm *remote, struct msghdr *msg,
600 static bool hvs_dgram_allow(u32 cid, u32 port)
605 static int hvs_update_recv_data(struct hvsock *hvs)
607 struct hvs_recv_buf *recv_buf;
610 recv_buf = (struct hvs_recv_buf *)(hvs->recv_desc + 1);
611 payload_len = recv_buf->hdr.data_size;
613 if (payload_len > HVS_MTU_SIZE)
616 if (payload_len == 0)
617 hvs->vsk->peer_shutdown |= SEND_SHUTDOWN;
619 hvs->recv_data_len = payload_len;
620 hvs->recv_data_off = 0;
625 static ssize_t hvs_stream_dequeue(struct vsock_sock *vsk, struct msghdr *msg,
626 size_t len, int flags)
628 struct hvsock *hvs = vsk->trans;
629 bool need_refill = !hvs->recv_desc;
630 struct hvs_recv_buf *recv_buf;
634 if (flags & MSG_PEEK)
638 hvs->recv_desc = hv_pkt_iter_first(hvs->chan);
639 ret = hvs_update_recv_data(hvs);
644 recv_buf = (struct hvs_recv_buf *)(hvs->recv_desc + 1);
645 to_read = min_t(u32, len, hvs->recv_data_len);
646 ret = memcpy_to_msg(msg, recv_buf->data + hvs->recv_data_off, to_read);
650 hvs->recv_data_len -= to_read;
651 if (hvs->recv_data_len == 0) {
652 hvs->recv_desc = hv_pkt_iter_next(hvs->chan, hvs->recv_desc);
653 if (hvs->recv_desc) {
654 ret = hvs_update_recv_data(hvs);
659 hvs->recv_data_off += to_read;
665 static ssize_t hvs_stream_enqueue(struct vsock_sock *vsk, struct msghdr *msg,
668 struct hvsock *hvs = vsk->trans;
669 struct vmbus_channel *chan = hvs->chan;
670 struct hvs_send_buf *send_buf;
671 ssize_t to_write, max_writable;
673 ssize_t bytes_written = 0;
675 BUILD_BUG_ON(sizeof(*send_buf) != HV_HYP_PAGE_SIZE);
677 send_buf = kmalloc(sizeof(*send_buf), GFP_KERNEL);
681 /* Reader(s) could be draining data from the channel as we write.
682 * Maximize bandwidth, by iterating until the channel is found to be
686 max_writable = hvs_channel_writable_bytes(chan);
689 to_write = min_t(ssize_t, len, max_writable);
690 to_write = min_t(ssize_t, to_write, HVS_SEND_BUF_SIZE);
691 /* memcpy_from_msg is safe for loop as it advances the offsets
692 * within the message iterator.
694 ret = memcpy_from_msg(send_buf->data, msg, to_write);
698 ret = hvs_send_data(hvs->chan, send_buf, to_write);
702 bytes_written += to_write;
706 /* If any data has been sent, return that */
713 static s64 hvs_stream_has_data(struct vsock_sock *vsk)
715 struct hvsock *hvs = vsk->trans;
718 if (hvs->recv_data_len > 0)
721 switch (hvs_channel_readable_payload(hvs->chan)) {
726 vsk->peer_shutdown |= SEND_SHUTDOWN;
737 static s64 hvs_stream_has_space(struct vsock_sock *vsk)
739 struct hvsock *hvs = vsk->trans;
741 return hvs_channel_writable_bytes(hvs->chan);
744 static u64 hvs_stream_rcvhiwat(struct vsock_sock *vsk)
746 return HVS_MTU_SIZE + 1;
749 static bool hvs_stream_is_active(struct vsock_sock *vsk)
751 struct hvsock *hvs = vsk->trans;
753 return hvs->chan != NULL;
756 static bool hvs_stream_allow(u32 cid, u32 port)
758 /* The host's port range [MIN_HOST_EPHEMERAL_PORT, 0xFFFFFFFF) is
759 * reserved as ephemeral ports, which are used as the host's ports
760 * when the host initiates connections.
762 * Perform this check in the guest so an immediate error is produced
763 * instead of a timeout.
765 if (port > MAX_HOST_LISTEN_PORT)
768 if (cid == VMADDR_CID_HOST)
775 int hvs_notify_poll_in(struct vsock_sock *vsk, size_t target, bool *readable)
777 struct hvsock *hvs = vsk->trans;
779 *readable = hvs_channel_readable(hvs->chan);
784 int hvs_notify_poll_out(struct vsock_sock *vsk, size_t target, bool *writable)
786 *writable = hvs_stream_has_space(vsk) > 0;
792 int hvs_notify_recv_init(struct vsock_sock *vsk, size_t target,
793 struct vsock_transport_recv_notify_data *d)
799 int hvs_notify_recv_pre_block(struct vsock_sock *vsk, size_t target,
800 struct vsock_transport_recv_notify_data *d)
806 int hvs_notify_recv_pre_dequeue(struct vsock_sock *vsk, size_t target,
807 struct vsock_transport_recv_notify_data *d)
813 int hvs_notify_recv_post_dequeue(struct vsock_sock *vsk, size_t target,
814 ssize_t copied, bool data_read,
815 struct vsock_transport_recv_notify_data *d)
821 int hvs_notify_send_init(struct vsock_sock *vsk,
822 struct vsock_transport_send_notify_data *d)
828 int hvs_notify_send_pre_block(struct vsock_sock *vsk,
829 struct vsock_transport_send_notify_data *d)
835 int hvs_notify_send_pre_enqueue(struct vsock_sock *vsk,
836 struct vsock_transport_send_notify_data *d)
842 int hvs_notify_send_post_enqueue(struct vsock_sock *vsk, ssize_t written,
843 struct vsock_transport_send_notify_data *d)
848 static void hvs_set_buffer_size(struct vsock_sock *vsk, u64 val)
853 static void hvs_set_min_buffer_size(struct vsock_sock *vsk, u64 val)
858 static void hvs_set_max_buffer_size(struct vsock_sock *vsk, u64 val)
863 static u64 hvs_get_buffer_size(struct vsock_sock *vsk)
868 static u64 hvs_get_min_buffer_size(struct vsock_sock *vsk)
873 static u64 hvs_get_max_buffer_size(struct vsock_sock *vsk)
878 static struct vsock_transport hvs_transport = {
879 .get_local_cid = hvs_get_local_cid,
881 .init = hvs_sock_init,
882 .destruct = hvs_destruct,
883 .release = hvs_release,
884 .connect = hvs_connect,
885 .shutdown = hvs_shutdown,
887 .dgram_bind = hvs_dgram_bind,
888 .dgram_dequeue = hvs_dgram_dequeue,
889 .dgram_enqueue = hvs_dgram_enqueue,
890 .dgram_allow = hvs_dgram_allow,
892 .stream_dequeue = hvs_stream_dequeue,
893 .stream_enqueue = hvs_stream_enqueue,
894 .stream_has_data = hvs_stream_has_data,
895 .stream_has_space = hvs_stream_has_space,
896 .stream_rcvhiwat = hvs_stream_rcvhiwat,
897 .stream_is_active = hvs_stream_is_active,
898 .stream_allow = hvs_stream_allow,
900 .notify_poll_in = hvs_notify_poll_in,
901 .notify_poll_out = hvs_notify_poll_out,
902 .notify_recv_init = hvs_notify_recv_init,
903 .notify_recv_pre_block = hvs_notify_recv_pre_block,
904 .notify_recv_pre_dequeue = hvs_notify_recv_pre_dequeue,
905 .notify_recv_post_dequeue = hvs_notify_recv_post_dequeue,
906 .notify_send_init = hvs_notify_send_init,
907 .notify_send_pre_block = hvs_notify_send_pre_block,
908 .notify_send_pre_enqueue = hvs_notify_send_pre_enqueue,
909 .notify_send_post_enqueue = hvs_notify_send_post_enqueue,
911 .set_buffer_size = hvs_set_buffer_size,
912 .set_min_buffer_size = hvs_set_min_buffer_size,
913 .set_max_buffer_size = hvs_set_max_buffer_size,
914 .get_buffer_size = hvs_get_buffer_size,
915 .get_min_buffer_size = hvs_get_min_buffer_size,
916 .get_max_buffer_size = hvs_get_max_buffer_size,
919 static int hvs_probe(struct hv_device *hdev,
920 const struct hv_vmbus_device_id *dev_id)
922 struct vmbus_channel *chan = hdev->channel;
924 hvs_open_connection(chan);
926 /* Always return success to suppress the unnecessary error message
927 * in vmbus_probe(): on error the host will rescind the device in
928 * 30 seconds and we can do cleanup at that time in
929 * vmbus_onoffer_rescind().
934 static int hvs_remove(struct hv_device *hdev)
936 struct vmbus_channel *chan = hdev->channel;
943 /* This isn't really used. See vmbus_match() and vmbus_probe() */
944 static const struct hv_vmbus_device_id id_table[] = {
948 static struct hv_driver hvs_drv = {
951 .id_table = id_table,
953 .remove = hvs_remove,
956 static int __init hvs_init(void)
960 if (vmbus_proto_version < VERSION_WIN10)
963 ret = vmbus_driver_register(&hvs_drv);
967 ret = vsock_core_init(&hvs_transport);
969 vmbus_driver_unregister(&hvs_drv);
976 static void __exit hvs_exit(void)
979 vmbus_driver_unregister(&hvs_drv);
982 module_init(hvs_init);
983 module_exit(hvs_exit);
985 MODULE_DESCRIPTION("Hyper-V Sockets");
986 MODULE_VERSION("1.0.0");
987 MODULE_LICENSE("GPL");
988 MODULE_ALIAS_NETPROTO(PF_VSOCK);