* * *
This series of VSOCK linux upstreaming patches include latest udpate from
VMware.
Summary of changes:
- Add include/linux/socket.h for AF_VSOCK.
- Cleanup some comments.
- Cleanup makefiles.
* * *
In an effort to improve the out-of-the-box experience with Linux
kernels for VMware users, VMware is working on readying the Virtual
Machine Communication Interface (vmw_vmci) and VMCI Sockets (VSOCK)
(vmw_vsock) kernel modules for inclusion in the Linux kernel. The
purpose of this post is to acquire feedback on the vmw_vsock kernel
module. The vmw_vmci kernel module has been presented in an early post.
* * *
VMCI Sockets allows virtual machines to communicate with host kernel
modules and the VMware hypervisors. VMCI Sockets kernel module has
dependency on VMCI kernel module. User level applications both in
a virtual machine and on the host can use vmw_vmci through VMCI
Sockets API which facilitates fast and efficient communication
between guest virtual machines and their host. A socket
address family designed to be compatible with UDP and TCP at the
interface level. Today, VMCI and VMCI Sockets are used by the VMware
shared folders (HGFS) and various VMware Tools components inside the
guest for zero-config, network-less access to VMware host services. In
addition to this, VMware's users are using VMCI Sockets for various
applications, where network access of the virtual machine is
restricted or non-existent. Examples of this are VMs communicating
with device proxies for proprietary hardware running as host
applications and automated testing of applications running within
virtual machines.
The VMware VMCI Sockets are similar to other socket types, like
Berkeley UNIX socket interface. The VMCI sockets module supports
both connection-oriented stream sockets like TCP, and connectionless
datagram sockets like UDP. The VSOCK protocol family is defined as
"AF_VSOCK" and the socket operations split for SOCK_DGRAM and
SOCK_STREAM.
For additional information about the use of VMCI and in particular
VMCI Sockets, please refer to the VMCI Socket Programming Guide
available at https://www.vmware.com/support/developer/vmci-sdk/.
---
George Zhang (6):
VSOCK: vsock protocol implementation.
VSOCK: vsock address implementaion.
VSOCK: notification implementation.
VSOCK: statistics implementation.
VSOCK: utility functions.
VSOCK: header and config files.
include/linux/socket.h | 4
net/Kconfig | 1
net/Makefile | 1
net/vmw_vsock/Kconfig | 14
net/vmw_vsock/Makefile | 4
net/vmw_vsock/af_vsock.c | 4165 +++++++++++++++++++++++++++++++++++
net/vmw_vsock/af_vsock.h | 179 ++
net/vmw_vsock/notify.c | 984 ++++++++
net/vmw_vsock/notify.h | 130 +
net/vmw_vsock/notify_qstate.c | 627 +++++
net/vmw_vsock/stats.c | 37
net/vmw_vsock/stats.h | 219 ++
net/vmw_vsock/util.c | 626 +++++
net/vmw_vsock/util.h | 312 +++
net/vmw_vsock/vmci_sockets.h | 517 ++++
net/vmw_vsock/vmci_sockets_packet.h | 106 +
net/vmw_vsock/vsock_addr.c | 264 ++
net/vmw_vsock/vsock_addr.h | 40
net/vmw_vsock/vsock_common.h | 130 +
net/vmw_vsock/vsock_packet.h | 131 +
net/vmw_vsock/vsock_version.h | 29
21 files changed, 8519 insertions(+), 1 deletions(-)
create mode 100644 net/vmw_vsock/Kconfig
create mode 100644 net/vmw_vsock/Makefile
create mode 100644 net/vmw_vsock/af_vsock.c
create mode 100644 net/vmw_vsock/af_vsock.h
create mode 100644 net/vmw_vsock/notify.c
create mode 100644 net/vmw_vsock/notify.h
create mode 100644 net/vmw_vsock/notify_qstate.c
create mode 100644 net/vmw_vsock/stats.c
create mode 100644 net/vmw_vsock/stats.h
create mode 100644 net/vmw_vsock/util.c
create mode 100644 net/vmw_vsock/util.h
create mode 100644 net/vmw_vsock/vmci_sockets.h
create mode 100644 net/vmw_vsock/vmci_sockets_packet.h
create mode 100644 net/vmw_vsock/vsock_addr.c
create mode 100644 net/vmw_vsock/vsock_addr.h
create mode 100644 net/vmw_vsock/vsock_common.h
create mode 100644 net/vmw_vsock/vsock_packet.h
create mode 100644 net/vmw_vsock/vsock_version.h
--
Signature
VSOCK linux socket module for VMCI Sockets protocol family.
Signed-off-by: George Zhang <[email protected]>
---
net/vmw_vsock/af_vsock.c | 4165 ++++++++++++++++++++++++++++++++++++++++++++++
net/vmw_vsock/af_vsock.h | 179 ++
2 files changed, 4344 insertions(+), 0 deletions(-)
create mode 100644 net/vmw_vsock/af_vsock.c
create mode 100644 net/vmw_vsock/af_vsock.h
diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
new file mode 100644
index 0000000..00c2050
--- /dev/null
+++ b/net/vmw_vsock/af_vsock.c
@@ -0,0 +1,4165 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * af_vsock.c --
+ *
+ * Linux socket module for the VMCI Sockets protocol family.
+ */
+
+/*
+ * Implementation notes:
+ *
+ * - There are two kinds of sockets: those created by user action (such as
+ * calling socket(2)) and those created by incoming connection request packets.
+ *
+ * - There are two "global" tables, one for bound sockets (sockets that have
+ * specified an address that they are responsible for) and one for connected
+ * sockets (sockets that have established a connection with another socket).
+ * These tables are "global" in that all sockets on the system are placed
+ * within them. - Note, though, that the bound table contains an extra entry
+ * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in
+ * that list. The bound table is used solely for lookup of sockets when packets
+ * are received and that's not necessary for SOCK_DGRAM sockets since we create
+ * a datagram handle for each and need not perform a lookup. Keeping SOCK_DGRAM
+ * sockets out of the bound hash buckets will reduce the chance of collisions
+ * when looking for SOCK_STREAM sockets and prevents us from having to check the
+ * socket type in the hash table lookups.
+ *
+ * - Sockets created by user action will either be "client" sockets that
+ * initiate a connection or "server" sockets that listen for connections; we do
+ * not support simultaneous connects (two "client" sockets connecting).
+ *
+ * - "Server" sockets are referred to as listener sockets throughout this
+ * implementation because they are in the SS_LISTEN state. When a connection
+ * request is received (the second kind of socket mentioned above), we create a
+ * new socket and refer to it as a pending socket. These pending sockets are
+ * placed on the pending connection list of the listener socket. When future
+ * packets are received for the address the listener socket is bound to, we
+ * check if the source of the packet is from one that has an existing pending
+ * connection. If it does, we process the packet for the pending socket. When
+ * that socket reaches the connected state, it is removed from the listener
+ * socket's pending list and enqueued in the listener socket's accept queue.
+ * Callers of accept(2) will accept connected sockets from the listener socket's
+ * accept queue. If the socket cannot be accepted for some reason then it is
+ * marked rejected. Once the connection is accepted, it is owned by the user
+ * process and the responsibility for cleanup falls with that user process.
+ *
+ * - It is possible that these pending sockets will never reach the connected
+ * state; in fact, we may never receive another packet after the connection
+ * request. Because of this, we must schedule a cleanup function to run in the
+ * future, after some amount of time passes where a connection should have been
+ * established. This function ensures that the socket is off all lists so it
+ * cannot be retrieved, then drops all references to the socket so it is cleaned
+ * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this
+ * function will also cleanup rejected sockets, those that reach the connected
+ * state but leave it before they have been accepted.
+ *
+ * - Sockets created by user action will be cleaned up when the user process
+ * calls close(2), causing our release implementation to be called. Our release
+ * implementation will perform some cleanup then drop the last reference so our
+ * sk_destruct implementation is invoked. Our sk_destruct implementation will
+ * perform additional cleanup that's common for both types of sockets.
+ *
+ * - A socket's reference count is what ensures that the structure won't be
+ * freed. Each entry in a list (such as the "global" bound and connected tables
+ * and the listener socket's pending list and connected queue) ensures a
+ * reference. When we defer work until process context and pass a socket as our
+ * argument, we must ensure the reference count is increased to ensure the
+ * socket isn't freed before the function is run; the deferred function will
+ * then drop the reference.
+ */
+
+#include <linux/types.h>
+
+#define EXPORT_SYMTAB
+#include <linux/kmod.h>
+#include <linux/socket.h>
+#include <linux/net.h>
+#include <linux/skbuff.h>
+#include <linux/miscdevice.h>
+#include <linux/poll.h>
+#include <linux/smp.h>
+#include <linux/bitops.h>
+#include <linux/list.h>
+#include <linux/wait.h>
+#include <linux/init.h>
+#include <linux/io.h>
+
+#include <linux/module.h>
+#include <linux/unistd.h>
+#include <linux/stddef.h> /* for NULL */
+#include <net/sock.h>
+#include <linux/kernel.h>
+#include <linux/workqueue.h>
+#include <linux/mutex.h>
+/*
+ * Include linux/cred.h via linux/sched.h - it is not nice, but as cpp does not
+ * have #ifexist...
+ */
+#include <linux/sched.h>
+
+#if !defined(cap_set_full)
+/* cap_set_full was removed in kernel version 3.0-rc4. */
+#define cap_set_full(_c) do { (_c) = CAP_FULL_SET; } while (0)
+#endif
+
+#include "af_vsock.h"
+#include "stats.h"
+#include "util.h"
+#include "vsock_version.h"
+
+/*
+ * All kernels above 2.6.33 have the kern parameter for the create call in
+ * struct net_proto_family.
+ */
+
+#define VSOCK_INVALID_FAMILY NPROTO
+#define VSOCK_AF_IS_REGISTERED(val) ((val) >= 0 && (val) < NPROTO)
+
+/*
+ * Prototypes
+ */
+
+/* Internal functions. */
+static bool vsock_vmci_proto_to_notify_struct(struct sock *sk,
+ vsock_proto_version * proto,
+ bool old_pkt_proto);
+static int vsock_vmci_recv_dgram_cb(void *data, struct vmci_datagram *dg);
+static int vsock_vmci_recv_stream_cb(void *data, struct vmci_datagram *dg);
+static void vsock_vmci_peer_attach_cb(vmci_id sub_id,
+ const struct vmci_event_data *ed,
+ void *client_data);
+static void vsock_vmci_peer_detach_cb(vmci_id sub_id,
+ const struct vmci_event_data *ed,
+ void *client_data);
+static void vsock_vmci_recv_pkt_work(struct work_struct *work);
+static int vsock_vmci_recv_listen(struct sock *sk, vsock_packet *pkt);
+static int vsock_vmci_recv_connecting_server(struct sock *sk,
+ struct sock *pending,
+ vsock_packet *pkt);
+static int vsock_vmci_recv_connecting_client(struct sock *sk,
+ vsock_packet *pkt);
+static int vsock_vmci_recv_connecting_client_negotiate(struct sock *sk,
+ vsock_packet *pkt);
+static int vsock_vmci_recv_connecting_client_invalid(struct sock *sk,
+ vsock_packet *pkt);
+static int vsock_vmci_recv_connected(struct sock *sk, vsock_packet *pkt);
+static int __vsock_vmci_bind(struct sock *sk, struct sockaddr_vm *addr);
+static struct sock *__vsock_vmci_create(struct net *net,
+ struct socket *sock,
+ struct sock *parent, gfp_t priority,
+ unsigned short type);
+static int vsock_vmci_register_with_vmci(void);
+static void vsock_vmci_unregister_with_vmci(void);
+
+/* Socket operations. */
+static void vsock_vmci_sk_destruct(struct sock *sk);
+static int vsock_vmci_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
+static int vsock_vmci_release(struct socket *sock);
+static int vsock_vmci_bind(struct socket *sock,
+ struct sockaddr *addr, int addr_len);
+static int vsock_vmci_dgram_connect(struct socket *sock,
+ struct sockaddr *addr, int addr_len,
+ int flags);
+static int vsock_vmci_stream_connect(struct socket *sock, struct sockaddr *addr,
+ int addr_len, int flags);
+static int vsock_vmci_accept(struct socket *sock, struct socket *newsock,
+ int flags);
+static int vsock_vmci_getname(struct socket *sock, struct sockaddr *addr,
+ int *addr_len, int peer);
+static unsigned int vsock_vmci_poll(struct file *file, struct socket *sock,
+ poll_table *wait);
+static int vsock_vmci_listen(struct socket *sock, int backlog);
+static int vsock_vmci_shutdown(struct socket *sock, int mode);
+
+typedef unsigned int vsock_setsockopt_len_type;
+static int vsock_vmci_stream_setsockopt(struct socket *sock, int level,
+ int optname, char __user *optval,
+ vsock_setsockopt_len_type optlen);
+
+static int vsock_vmci_stream_getsockopt(struct socket *sock, int level,
+ int optname, char __user *optval,
+ int __user *optlen);
+
+static int vsock_vmci_dgram_sendmsg(struct kiocb *kiocb,
+ struct socket *sock, struct msghdr *msg,
+ size_t len);
+static int vsock_vmci_dgram_recvmsg(struct kiocb *kiocb, struct socket *sock,
+ struct msghdr *msg, size_t len, int flags);
+static int vsock_vmci_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
+ struct msghdr *msg, size_t len);
+static int vsock_vmci_stream_recvmsg(struct kiocb *kiocb, struct socket *sock,
+ struct msghdr *msg, size_t len, int flags);
+
+static int vsock_vmci_create(struct net *net,
+ struct socket *sock, int protocol, int kern);
+
+/*
+ * Variables.
+ */
+
+/* Protocol family. */
+static struct proto vsock_vmci_proto = {
+ .name = "AF_VMCI",
+ /* Added in 2.6.10. */
+ .owner = THIS_MODULE,
+ /*
+ * From 2.6.9 until 2.6.11, these address families called
+ * sk_alloc_slab() and the allocated slab was assigned to the slab
+ * variable in the proto struct and was created of size slab_obj_size.
+ * As of 2.6.12 and later, this slab allocation was moved into
+ * proto_register() and only done if you specified a non-zero value for
+ * the second argument (alloc_slab); the size of the slab element was
+ * changed to obj_size.
+ */
+ .obj_size = sizeof(vsock_vmci_sock),
+};
+
+static const struct net_proto_family vsock_vmci_family_ops = {
+ .family = AF_VSOCK,
+ .create = vsock_vmci_create,
+ .owner = THIS_MODULE,
+};
+
+/* Socket operations, split for DGRAM and STREAM sockets. */
+static const struct proto_ops vsock_vmci_dgram_ops = {
+ .family = PF_VSOCK,
+ .owner = THIS_MODULE,
+ .release = vsock_vmci_release,
+ .bind = vsock_vmci_bind,
+ .connect = vsock_vmci_dgram_connect,
+ .socketpair = sock_no_socketpair,
+ .accept = sock_no_accept,
+ .getname = vsock_vmci_getname,
+ .poll = vsock_vmci_poll,
+ .ioctl = sock_no_ioctl,
+ .listen = sock_no_listen,
+ .shutdown = vsock_vmci_shutdown,
+ .setsockopt = sock_no_setsockopt,
+ .getsockopt = sock_no_getsockopt,
+ .sendmsg = vsock_vmci_dgram_sendmsg,
+ .recvmsg = vsock_vmci_dgram_recvmsg,
+ .mmap = sock_no_mmap,
+ .sendpage = sock_no_sendpage,
+};
+
+static const struct proto_ops vsock_vmci_stream_ops = {
+ .family = PF_VSOCK,
+ .owner = THIS_MODULE,
+ .release = vsock_vmci_release,
+ .bind = vsock_vmci_bind,
+ .connect = vsock_vmci_stream_connect,
+ .socketpair = sock_no_socketpair,
+ .accept = vsock_vmci_accept,
+ .getname = vsock_vmci_getname,
+ .poll = vsock_vmci_poll,
+ .ioctl = sock_no_ioctl,
+ .listen = vsock_vmci_listen,
+ .shutdown = vsock_vmci_shutdown,
+ .setsockopt = vsock_vmci_stream_setsockopt,
+ .getsockopt = vsock_vmci_stream_getsockopt,
+ .sendmsg = vsock_vmci_stream_sendmsg,
+ .recvmsg = vsock_vmci_stream_recvmsg,
+ .mmap = sock_no_mmap,
+ .sendpage = sock_no_sendpage,
+};
+
+typedef struct vsock_recv_pkt_info {
+ struct work_struct work;
+ struct sock *sk;
+ vsock_packet pkt;
+} vsock_recv_pkt_info;
+
+static struct vmci_handle vmci_stream_handle = { VMCI_INVALID_ID,
+ VMCI_INVALID_ID };
+
+static vmci_id qp_resumed_sub_id = VMCI_INVALID_ID;
+
+static int PROTOCOL_OVERRIDE = -1;
+
+/*
+ * Netperf benchmarks have shown significant throughput improvements when the
+ * QP size is bumped from 64k to 256k. These measurements were taken during the
+ * K/L.next timeframe. Give users better performance by default.
+ */
+#define VSOCK_DEFAULT_QP_SIZE_MIN 128
+#define VSOCK_DEFAULT_QP_SIZE 262144
+#define VSOCK_DEFAULT_QP_SIZE_MAX 262144
+
+/*
+ * The default peer timeout indicates how long we will wait for a peer response
+ * to a control message.
+ */
+#define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
+
+#define LOG_PACKET(_pkt)
+
+/**
+ * vsock_vmci_old_proto_override --
+ *
+ * Check to see if the user has asked us to override all sockets to use the
+ * vsock notify protocol.
+ *
+ * Results: true if there is a protocol override in effect. - old_pkt_proto is
+ * true the original protocol should be used. False if there is no override in
+ * effect.
+ *
+ * Side effects: None.
+ */
+
+static bool vsock_vmci_old_proto_override(bool *old_pkt_proto)
+{
+ if (PROTOCOL_OVERRIDE != -1) {
+ if (PROTOCOL_OVERRIDE == 0)
+ *old_pkt_proto = true;
+ else
+ *old_pkt_proto = false;
+
+ pr_info("Proto override in use.\n");
+ return true;
+ }
+
+ return false;
+}
+
+/*
+ * vsock_vmci_proto_to_notify_struct --
+ *
+ * Given a particular notify protocol version, setup the socket's notify struct
+ * correctly.
+ *
+ * Results: true on success, false otherwise.
+ *
+ * Side effects: None.
+ */
+
+static bool
+vsock_vmci_proto_to_notify_struct(struct sock *sk,
+ vsock_proto_version *proto,
+ bool old_pkt_proto)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ if (old_pkt_proto) {
+ if (*proto != VSOCK_PROTO_INVALID) {
+ pr_err("Can't set both an old and new protocol\n");
+ return false;
+ }
+ vsk->notify_ops = &vsock_vmci_notify_pkt_ops;
+ goto exit;
+ }
+
+ switch (*proto) {
+ case VSOCK_PROTO_PKT_ON_NOTIFY:
+ vsk->notify_ops = &vsock_vmci_notify_pkt_q_state_ops;
+ break;
+ default:
+ pr_err("Unknown notify protocol version\n");
+ return false;
+ }
+
+exit:
+ NOTIFYCALL(vsk, socket_init, sk);
+ return true;
+}
+
+/*
+ * vsock_vmci_new_proto_supported_versions
+ *
+ * Gets the supported REQUEST2/NEGOTIATE2 vsock protocol versions.
+ *
+ * Results: Either 1 specific protocol version (override mode) or
+ * VSOCK_PROTO_ALL_SUPPORTED.
+ *
+ * Side effects: None.
+ */
+
+static vsock_proto_version vsock_vmci_new_proto_supported_versions(void)
+{
+ if (PROTOCOL_OVERRIDE != -1)
+ return PROTOCOL_OVERRIDE;
+
+ return VSOCK_PROTO_ALL_SUPPORTED;
+}
+
+/*
+ * VSockSocket_Trusted --
+ *
+ * We allow two kinds of sockets to communicate with a restricted VM: 1)
+ * trusted sockets 2) sockets from applications running as the same user as the
+ * VM (this is only true for the host side and only when using hosted products)
+ *
+ * Results: true if trusted communication is allowed to peer_cid, false
+ * otherwise.
+ *
+ * Side effects: None.
+ */
+
+static bool vsock_vmci_trusted(vsock_vmci_sock *vsock, vmci_id peer_cid)
+{
+ return vsock->trusted ||
+ vmci_is_context_owner(peer_cid, vsock->owner->uid);
+}
+
+/*
+ * VSockSocket_AllowDgram --
+ *
+ * We allow sending datagrams to and receiving datagrams from a restricted VM
+ * only if it is trusted as described in vsock_vmci_trusted.
+ *
+ * Results: true if datagram communication is allowed to peer_cid, false
+ * otherwise.
+ *
+ * Side effects: None.
+ */
+
+static bool vsock_vmci_allow_dgram(vsock_vmci_sock *vsock, vmci_id peer_cid)
+{
+ if (vsock->cached_peer != peer_cid) {
+ vsock->cached_peer = peer_cid;
+ if (!vsock_vmci_trusted(vsock, peer_cid) &&
+ (vmci_context_get_priv_flags(peer_cid) &
+ VMCI_PRIVILEGE_FLAG_RESTRICTED)) {
+ vsock->cached_peer_allow_dgram = false;
+ } else {
+ vsock->cached_peer_allow_dgram = true;
+ }
+ }
+
+ return vsock->cached_peer_allow_dgram;
+}
+
+/*
+ * vmci_sock_get_local_c_id --
+ *
+ * Kernel interface that allows external kernel modules to get the current VMCI
+ * context id. This version of the function is exported to kernel clients and
+ * should not change.
+ *
+ * Results: The context id on success, a negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+int vmci_sock_get_local_c_id(void)
+{
+ /* FIXME: needed? */
+ return vmci_get_context_id();
+}
+EXPORT_SYMBOL_GPL(vmci_sock_get_local_c_id);
+
+/*
+ * Helper functions.
+ */
+
+/*
+ * vsock_vmci_queue_pair_alloc --
+ *
+ * Allocates or attaches to a queue pair. Tries to register with trusted status
+ * if requested but does not fail if the queuepair could not be allocate as
+ * trusted (running in the guest)
+ *
+ * Results: 0 on success. A VSock error on error.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_queue_pair_alloc(struct vmci_qp **qpair,
+ struct vmci_handle *handle,
+ u64 produce_size,
+ u64 consume_size,
+ vmci_id peer, u32 flags, bool trusted)
+{
+ int err = 0;
+
+ if (trusted) {
+ /*
+ * Try to allocate our queue pair as trusted. This will only
+ * work if vsock is running in the host.
+ */
+
+ err = vmci_qpair_alloc(qpair, handle, produce_size,
+ consume_size,
+ peer, flags,
+ VMCI_PRIVILEGE_FLAG_TRUSTED);
+ if (err != VMCI_ERROR_NO_ACCESS)
+ goto out;
+
+ }
+
+ err = vmci_qpair_alloc(qpair, handle, produce_size, consume_size,
+ peer, flags, VMCI_NO_PRIVILEGE_FLAGS);
+out:
+ if (err < 0) {
+ pr_err("Could not attach to queue pair with %d\n",
+ err);
+ err = vsock_vmci_error_to_vsock_error(err);
+ }
+
+ return err;
+}
+
+/*
+ * vsock_vmci_datagram_create_hnd --
+ *
+ * Creates a datagram handle. Tries to register with trusted status but does
+ * not fail if the handler could not be allocated as trusted (running in the
+ * guest).
+ *
+ * Results: 0 on success. A VMCI error on error.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_datagram_create_hnd(vmci_id resource_id,
+ u32 flags,
+ vmci_datagram_recv_cb recv_cb,
+ void *client_data,
+ struct vmci_handle *out_handle)
+{
+ int err = 0;
+
+ /*
+ * Try to allocate our datagram handler as trusted. This will only work
+ * if vsock is running in the host.
+ */
+
+ err = vmci_datagram_create_handle_priv(resource_id, flags,
+ VMCI_PRIVILEGE_FLAG_TRUSTED,
+ recv_cb,
+ client_data, out_handle);
+
+ if (err == VMCI_ERROR_NO_ACCESS)
+ err = vmci_datagram_create_handle(resource_id, flags,
+ recv_cb, client_data,
+ out_handle);
+
+ return err;
+}
+
+/*
+ * vsock_vmci_recv_dgram_cb --
+ *
+ * VMCI Datagram receive callback. This function is used specifically for
+ * SOCK_DGRAM sockets.
+ *
+ * This is invoked as part of a tasklet that's scheduled when the VMCI
+ * interrupt fires. This is run in bottom-half context and if it ever needs to
+ * sleep it should defer that work to a work queue.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: An sk_buff is created and queued with this socket.
+ */
+
+static int vsock_vmci_recv_dgram_cb(void *data, struct vmci_datagram *dg)
+{
+ struct sock *sk;
+ size_t size;
+ struct sk_buff *skb;
+ vsock_vmci_sock *vsk;
+
+ ASSERT(dg->payload_size <= VMCI_MAX_DG_PAYLOAD_SIZE);
+
+ sk = (struct sock *)data;
+
+ /*
+ * This handler is privileged when this module is running on the host.
+ * We will get datagrams from all endpoints (even VMs that are in a
+ * restricted context). If we get one from a restricted context then
+ * the destination socket must be trusted.
+ *
+ * NOTE: We access the socket struct without holding the lock here.
+ * This is ok because the field we are interested is never modified
+ * outside of the create and destruct socket functions.
+ */
+ vsk = vsock_sk(sk);
+ if (!vsock_vmci_allow_dgram(vsk, VMCI_HANDLE_TO_CONTEXT_ID(dg->src)))
+ return VMCI_ERROR_NO_ACCESS;
+
+ size = VMCI_DG_SIZE(dg);
+
+ /*
+ * Attach the packet to the socket's receive queue as an sk_buff.
+ */
+ skb = alloc_skb(size, GFP_ATOMIC);
+ if (skb) {
+ /* sk_receive_skb() will do a sock_put(), so hold here. */
+ sock_hold(sk);
+ skb_put(skb, size);
+ memcpy(skb->data, dg, size);
+ sk_receive_skb(sk, skb, 0);
+ }
+
+ return VMCI_SUCCESS;
+}
+
+/*
+ * vsock_vmci_recv_stream_cb --
+ *
+ * VMCI stream receive callback for control datagrams. This function is used
+ * specifically for SOCK_STREAM sockets.
+ *
+ * This is invoked as part of a tasklet that's scheduled when the VMCI
+ * interrupt fires. This is run in bottom-half context but it defers most of
+ * its work to the packet handling work queue.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_recv_stream_cb(void *data, struct vmci_datagram *dg)
+{
+ struct sock *sk;
+ struct sockaddr_vm dst;
+ struct sockaddr_vm src;
+ vsock_packet *pkt;
+ vsock_vmci_sock *vsk;
+ bool bh_process_pkt;
+ int err;
+
+ ASSERT(dg->payload_size <= VMCI_MAX_DG_PAYLOAD_SIZE);
+
+ sk = NULL;
+ err = VMCI_SUCCESS;
+ bh_process_pkt = false;
+
+ /*
+ * Ignore incoming packets from contexts without sockets, or resources
+ * that aren't vsock implementations.
+ */
+
+ if (!vsock_addr_socket_context_stream
+ (VMCI_HANDLE_TO_CONTEXT_ID(dg->src))
+ || VSOCK_PACKET_RID != VMCI_HANDLE_TO_RESOURCE_ID(dg->src)) {
+ return VMCI_ERROR_NO_ACCESS;
+ }
+
+ if (VMCI_DG_SIZE(dg) < sizeof *pkt)
+ /* Drop datagrams that do not contain full VSock packets. */
+ return VMCI_ERROR_INVALID_ARGS;
+
+ pkt = (vsock_packet *) dg;
+
+ LOG_PACKET(pkt);
+
+ /*
+ * Find the socket that should handle this packet. First we look for a
+ * connected socket and if there is none we look for a socket bound to
+ * the destintation address.
+ */
+ vsock_addr_init(&src, VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src),
+ pkt->src_port);
+
+ vsock_addr_init(&dst, VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.dst),
+ pkt->dst_port);
+
+ sk = vsock_vmci_find_connected_socket(&src, &dst);
+ if (!sk) {
+ sk = vsock_vmci_find_bound_socket(&dst);
+ if (!sk) {
+ /*
+ * We could not find a socket for this specified
+ * address. If this packet is a RST, we just drop it.
+ * If it is another packet, we send a RST. Note that
+ * we do not send a RST reply to RSTs so that we do not
+ * continually send RSTs between two endpoints.
+ *
+ * Note that since this is a reply, dst is src and src
+ * is dst.
+ */
+ if (VSOCK_SEND_RESET_BH(&dst, &src, pkt) < 0)
+ pr_err("unable to send reset.\n");
+
+ err = VMCI_ERROR_NOT_FOUND;
+ goto out;
+ }
+ }
+
+ /*
+ * If the received packet type is beyond all types known to this
+ * implementation, reply with an invalid message. Hopefully this will
+ * help when implementing backwards compatibility in the future.
+ */
+ if (pkt->type >= VSOCK_PACKET_TYPE_MAX) {
+ VSOCK_SEND_INVALID_BH(&dst, &src);
+ err = VMCI_ERROR_INVALID_ARGS;
+ goto out;
+ }
+
+ /*
+ * This handler is privileged when this module is running on the host.
+ * We will get datagram connect requests from all endpoints (even VMs
+ * that are in a restricted context). If we get one from a restricted
+ * context then the destination socket must be trusted.
+ *
+ * NOTE: We access the socket struct without holding the lock here.
+ * This is ok because the field we are interested is never modified
+ * outside of the create and destruct socket functions.
+ */
+ vsk = vsock_sk(sk);
+ if (!vsock_vmci_allow_dgram
+ (vsk, VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src))) {
+ err = VMCI_ERROR_NO_ACCESS;
+ goto out;
+ }
+
+ /*
+ * We do most everything in a work queue, but let's fast path the
+ * notification of reads and writes to help data transfer performance.
+ * We can only do this if there is no process context code executing
+ * for this socket since that may change the state.
+ */
+ bh_lock_sock(sk);
+
+ if (!sock_owned_by_user(sk) && sk->sk_state == SS_CONNECTED)
+ NOTIFYCALL(vsk, handle_notify_pkt, sk, pkt, true, &dst, &src,
+ &bh_process_pkt);
+
+ bh_unlock_sock(sk);
+
+ if (!bh_process_pkt) {
+ vsock_recv_pkt_info *recv_pkt_info;
+
+ recv_pkt_info = kmalloc(sizeof *recv_pkt_info, GFP_ATOMIC);
+ if (!recv_pkt_info) {
+ if (VSOCK_SEND_RESET_BH(&dst, &src, pkt) < 0)
+ pr_err("unable to send reset\n");
+
+ err = VMCI_ERROR_NO_MEM;
+ goto out;
+ }
+
+ recv_pkt_info->sk = sk;
+ memcpy(&recv_pkt_info->pkt, pkt, sizeof recv_pkt_info->pkt);
+ INIT_WORK(&recv_pkt_info->work, vsock_vmci_recv_pkt_work);
+
+ schedule_work(&recv_pkt_info->work);
+ /*
+ * Clear sk so that the reference count incremented by one of
+ * the Find functions above is not decremented below. We need
+ * that reference count for the packet handler we've scheduled
+ * to run.
+ */
+ sk = NULL;
+ }
+
+out:
+ if (sk)
+ sock_put(sk);
+
+ return err;
+}
+
+/*
+ * vsock_vmci_peer_attach_cb --
+ *
+ * Invoked when a peer attaches to a queue pair.
+ *
+ * Right now this does not do anything.
+ *
+ * Results: None.
+ *
+ * Side effects: May modify socket state and signal socket.
+ */
+
+static void vsock_vmci_peer_attach_cb(vmci_id sub_id,
+ const struct vmci_event_data *e_data,
+ void *client_data)
+{
+ struct sock *sk = client_data;
+ const struct vmci_event_payload_qp *e_payload;
+ vsock_vmci_sock *vsk;
+
+ e_payload = vmci_event_data_const_payload(e_data);
+
+ vsk = vsock_sk(sk);
+
+ /*
+ * We don't ask for delayed CBs when we subscribe to this event (we
+ * pass 0 as flags to vmci_event_subscribe()). VMCI makes no
+ * guarantees in that case about what context we might be running in,
+ * so it could be BH or process, blockable or non-blockable. So we
+ * need to account for all possible contexts here.
+ */
+ local_bh_disable();
+ bh_lock_sock(sk);
+
+ /*
+ * XXX This is lame, we should provide a way to lookup sockets by
+ * qp_handle.
+ */
+ if (VMCI_HANDLE_EQUAL(vsk->qp_handle, e_payload->handle)) {
+ /*
+ * XXX This doesn't do anything, but in the future we may want
+ * to set a flag here to verify the attach really did occur and
+ * we weren't just sent a datagram claiming it was.
+ */
+ goto out;
+ }
+
+out:
+ bh_unlock_sock(sk);
+ local_bh_enable();
+}
+
+/*
+ *
+ * vsock_vmci_handle_detach --
+ *
+ * Perform the work necessary when the peer has detached.
+ *
+ * Note that this assumes the socket lock is held.
+ *
+ * Results: None.
+ *
+ * Side effects: The socket's and its peer's shutdown mask will be set
+ * appropriately, and any callers waiting on this socket will be awoken.
+ */
+
+static void vsock_vmci_handle_detach(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+ if (!VMCI_HANDLE_INVALID(vsk->qp_handle)) {
+ sock_set_flag(sk, SOCK_DONE);
+
+ /*
+ * On a detach the peer will not be sending or receiving
+ * anymore.
+ */
+ vsk->peer_shutdown = SHUTDOWN_MASK;
+
+ /*
+ * We should not be sending anymore since the peer won't be
+ * there to receive, but we can still receive if there is data
+ * left in our consume queue.
+ */
+ if (vsock_vmci_stream_has_data(vsk) <= 0) {
+ if (sk->sk_state == SS_CONNECTING) {
+ /*
+ * The peer may detach from a queue pair while
+ * we are still in the connecting state, i.e.,
+ * if the peer VM is killed after attaching to
+ * a queue pair, but before we complete the
+ * handshake. In that case, we treat the detach
+ * event like a reset.
+ */
+
+ sk->sk_state = SS_UNCONNECTED;
+ sk->sk_err = ECONNRESET;
+ sk->sk_error_report(sk);
+ return;
+ }
+ sk->sk_state = SS_UNCONNECTED;
+ }
+ sk->sk_state_change(sk);
+ }
+}
+
+/*
+ * vsock_vmci_peer_detach_cb --
+ *
+ * Invoked when a peer detaches from a queue pair.
+ *
+ * Results: None.
+ *
+ * Side effects: May modify socket state and signal socket.
+ */
+
+static void vsock_vmci_peer_detach_cb(vmci_id sub_id,
+ const struct vmci_event_data *e_data,
+ void *client_data)
+{
+ struct sock *sk = client_data;
+ const struct vmci_event_payload_qp *e_payload;
+ vsock_vmci_sock *vsk;
+
+ e_payload = vmci_event_data_const_payload(e_data);
+ vsk = vsock_sk(sk);
+ if (VMCI_HANDLE_INVALID(e_payload->handle))
+ return;
+
+ /* Same rules for locking as for peer_attach_cb(). */
+ local_bh_disable();
+ bh_lock_sock(sk);
+
+ /*
+ * XXX This is lame, we should provide a way to lookup sockets by
+ * qp_handle.
+ */
+ if (VMCI_HANDLE_EQUAL(vsk->qp_handle, e_payload->handle))
+ vsock_vmci_handle_detach(sk);
+
+ bh_unlock_sock(sk);
+ local_bh_enable();
+}
+
+/*
+ * vsock_vmci_qp_resumed_cb --
+ *
+ * Invoked when a VM is resumed. We must mark all connected stream sockets as
+ * detached.
+ *
+ * Results: None.
+ *
+ * Side effects: May modify socket state and signal socket.
+ */
+
+static void vsock_vmci_qp_resumed_cb(vmci_id sub_id,
+ const struct vmci_event_data *e_data,
+ void *client_data)
+{
+ int i;
+
+ spin_lock_bh(&vsock_table_lock);
+
+ /*
+ * XXX This loop should probably be provided by util.{h,c}, but that's
+ * for another day.
+ */
+ for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
+ vsock_vmci_sock *vsk;
+
+ list_for_each_entry(vsk, &vsock_connected_table[i],
+ connected_table) {
+ struct sock *sk = sk_vsock(vsk);
+
+ /*
+ * XXX Technically this is racy but the resulting
+ * outcome from such a race is relatively harmless. My
+ * next change will be a fix to this.
+ */
+ vsock_vmci_handle_detach(sk);
+ }
+ }
+
+ spin_unlock_bh(&vsock_table_lock);
+}
+
+/*
+ * vsock_vmci_pending_work --
+ *
+ * Releases the resources for a pending socket if it has not reached the
+ * connected state and been accepted by a user process.
+ *
+ * Results: None.
+ *
+ * Side effects: The socket may be removed from the connected list and all its
+ * resources freed.
+ */
+
+static void vsock_vmci_pending_work(struct work_struct *work)
+{
+ struct sock *sk;
+ struct sock *listener;
+ vsock_vmci_sock *vsk;
+ bool cleanup;
+
+ vsk = container_of(work, vsock_vmci_sock, dwork.work);
+ sk = sk_vsock(vsk);
+ listener = vsk->listener;
+ cleanup = true;
+
+ lock_sock(listener);
+ lock_sock(sk);
+
+ /*
+ * The socket should be on the pending list or the accept queue, but
+ * not both. It's also possible that the socket isn't on either.
+ */
+ ASSERT((vsock_vmci_is_pending(sk) && !vsock_vmci_in_accept_queue(sk))
+ || (!vsock_vmci_is_pending(sk) && vsock_vmci_in_accept_queue(sk))
+ || (!vsock_vmci_is_pending(sk)
+ && !vsock_vmci_in_accept_queue(sk)));
+
+ if (vsock_vmci_is_pending(sk)) {
+ vsock_vmci_remove_pending(listener, sk);
+ } else if (!vsk->rejected) {
+ /*
+ * We are not on the pending list and accept() did not reject
+ * us, so we must have been accepted by our user process. We
+ * just need to drop our references to the sockets and be on
+ * our way.
+ */
+ cleanup = false;
+ goto out;
+ }
+
+ listener->sk_ack_backlog--;
+
+ /*
+ * We need to remove ourself from the global connected sockets list so
+ * incoming packets can't find this socket, and to reduce the reference
+ * count.
+ */
+ if (vsock_vmci_in_connected_table(sk))
+ vsock_vmci_remove_connected(sk);
+
+ sk->sk_state = SS_FREE;
+
+out:
+ release_sock(sk);
+ release_sock(listener);
+ if (cleanup)
+ sock_put(sk);
+
+ sock_put(sk);
+ sock_put(listener);
+}
+
+/*
+ * vsock_vmci_recv_pkt_work --
+ *
+ * Handles an incoming control packet for the provided socket. This is the
+ * state machine for our stream sockets.
+ *
+ * Results: None.
+ *
+ * Side effects: May set state and wakeup threads waiting for socket state to
+ * change.
+ */
+
+static void vsock_vmci_recv_pkt_work(struct work_struct *work)
+{
+ vsock_recv_pkt_info *recv_pkt_info;
+ vsock_packet *pkt;
+ struct sock *sk;
+
+ recv_pkt_info = container_of(work, vsock_recv_pkt_info, work);
+ sk = recv_pkt_info->sk;
+ pkt = &recv_pkt_info->pkt;
+
+ ASSERT(pkt->type < VSOCK_PACKET_TYPE_MAX);
+
+ lock_sock(sk);
+
+ switch (sk->sk_state) {
+ case SS_LISTEN:
+ vsock_vmci_recv_listen(sk, pkt);
+ break;
+ case SS_CONNECTING:
+ /*
+ * Processing of pending connections for servers goes through
+ * the listening socket, so see vsock_vmci_recv_listen() for
+ * that path.
+ */
+ vsock_vmci_recv_connecting_client(sk, pkt);
+ break;
+ case SS_CONNECTED:
+ vsock_vmci_recv_connected(sk, pkt);
+ break;
+ default:
+ /*
+ * Because this function does not run in the same context as
+ * vsock_vmci_recv_stream_cb it is possible that the socket has
+ * closed. We need to let the other side know or it could be
+ * sitting in a connect and hang forever. Send a reset to
+ * prevent that.
+ */
+ VSOCK_SEND_RESET(sk, pkt);
+ goto out;
+ }
+
+out:
+ release_sock(sk);
+ kfree(recv_pkt_info);
+ /*
+ * Release reference obtained in the stream callback when we fetched
+ * this socket out of the bound or connected list.
+ */
+ sock_put(sk);
+}
+
+/*
+ * vsock_vmci_recv_listen --
+ *
+ * Receives packets for sockets in the listen state.
+ *
+ * Note that this assumes the socket lock is held.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: A new socket may be created and a negotiate control packet is
+ * sent.
+ */
+
+static int vsock_vmci_recv_listen(struct sock *sk, vsock_packet *pkt)
+{
+ struct sock *pending;
+ vsock_vmci_sock *vpending;
+ int err;
+ u64 qp_size;
+ bool old_request = false;
+ bool old_pkt_proto = false;
+
+ ASSERT(sk->sk_state == SS_LISTEN);
+
+ err = 0;
+
+ /*
+ * Because we are in the listen state, we could be receiving a packet
+ * for ourself or any previous connection requests that we received.
+ * If it's the latter, we try to find a socket in our list of pending
+ * connections and, if we do, call the appropriate handler for the
+ * state that that socket is in. Otherwise we try to service the
+ * connection request.
+ */
+ pending = vsock_vmci_get_pending(sk, pkt);
+ if (pending) {
+ lock_sock(pending);
+ switch (pending->sk_state) {
+ case SS_CONNECTING:
+ err =
+ vsock_vmci_recv_connecting_server(sk, pending, pkt);
+ break;
+ default:
+ VSOCK_SEND_RESET(pending, pkt);
+ err = -EINVAL;
+ }
+
+ if (err < 0)
+ vsock_vmci_remove_pending(sk, pending);
+
+ release_sock(pending);
+ vsock_vmci_release_pending(pending);
+
+ return err;
+ }
+
+ /*
+ * The listen state only accepts connection requests. Reply with a
+ * reset unless we received a reset.
+ */
+
+ if (!(pkt->type == VSOCK_PACKET_TYPE_REQUEST ||
+ pkt->type == VSOCK_PACKET_TYPE_REQUEST2)) {
+ VSOCK_REPLY_RESET(pkt);
+ return -EINVAL;
+ }
+
+ if (pkt->u.size == 0) {
+ VSOCK_REPLY_RESET(pkt);
+ return -EINVAL;
+ }
+
+ /*
+ * If this socket can't accommodate this connection request, we send a
+ * reset. Otherwise we create and initialize a child socket and reply
+ * with a connection negotiation.
+ */
+ if (sk->sk_ack_backlog >= sk->sk_max_ack_backlog) {
+ VSOCK_REPLY_RESET(pkt);
+ return -ECONNREFUSED;
+ }
+
+ pending = __vsock_vmci_create(sock_net(sk), NULL, sk, GFP_KERNEL,
+ sk->sk_type);
+ if (!pending) {
+ VSOCK_SEND_RESET(sk, pkt);
+ return -ENOMEM;
+ }
+
+ vpending = vsock_sk(pending);
+ ASSERT(vsock_sk(sk)->local_addr.svm_port == pkt->dst_port);
+
+ vsock_addr_init(&vpending->local_addr,
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.dst), pkt->dst_port);
+ vsock_addr_init(&vpending->remote_addr,
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src), pkt->src_port);
+
+ /*
+ * If the proposed size fits within our min/max, accept it. Otherwise
+ * propose our own size.
+ */
+ if (pkt->u.size >= vpending->queue_pair_min_size &&
+ pkt->u.size <= vpending->queue_pair_max_size) {
+ qp_size = pkt->u.size;
+ } else {
+ qp_size = vpending->queue_pair_size;
+ }
+
+ /*
+ * Figure out if we are using old or new requests based on the
+ * overrides pkt types sent by our peer.
+ */
+ if (vsock_vmci_old_proto_override(&old_pkt_proto)) {
+ old_request = old_pkt_proto;
+ } else {
+ if (pkt->type == VSOCK_PACKET_TYPE_REQUEST)
+ old_request = true;
+ else if (pkt->type == VSOCK_PACKET_TYPE_REQUEST2)
+ old_request = false;
+
+ }
+
+ if (old_request) {
+ /* Handle a REQUEST (or override) */
+ vsock_proto_version version = VSOCK_PROTO_INVALID;
+ if (vsock_vmci_proto_to_notify_struct(pending, &version, true))
+ err = VSOCK_SEND_NEGOTIATE(pending, qp_size);
+ else
+ err = -EINVAL;
+
+ } else {
+ /* Handle a REQUEST2 (or override) */
+ int proto_int = pkt->proto;
+ int pos;
+ u16 active_proto_version = 0;
+
+ /*
+ * The list of possible protocols is the intersection of all
+ * protocols the client supports ... plus all the protocols we
+ * support.
+ */
+ proto_int &= vsock_vmci_new_proto_supported_versions();
+
+ /* We choose the highest possible protocol version and use that
+ * one. */
+ pos = fls(proto_int);
+ if (pos) {
+ active_proto_version = (1 << (pos - 1));
+ if (vsock_vmci_proto_to_notify_struct
+ (pending, &active_proto_version, false))
+ err =
+ VSOCK_SEND_NEGOTIATE2(pending, qp_size,
+ active_proto_version);
+ else
+ err = -EINVAL;
+
+ } else {
+ err = -EINVAL;
+ }
+ }
+
+ if (err < 0) {
+ VSOCK_SEND_RESET(sk, pkt);
+ sock_put(pending);
+ err = vsock_vmci_error_to_vsock_error(err);
+ goto out;
+ }
+
+ vsock_vmci_add_pending(sk, pending);
+ sk->sk_ack_backlog++;
+
+ pending->sk_state = SS_CONNECTING;
+ vpending->produce_size = vpending->consume_size = qp_size;
+ vpending->queue_pair_size = qp_size;
+
+ NOTIFYCALL(vpending, process_request, pending);
+
+ /*
+ * We might never receive another message for this socket and it's not
+ * connected to any process, so we have to ensure it gets cleaned up
+ * ourself. Our delayed work function will take care of that. Note
+ * that we do not ever cancel this function since we have few
+ * guarantees about its state when calling cancel_delayed_work().
+ * Instead we hold a reference on the socket for that function and make
+ * it capable of handling cases where it needs to do nothing but
+ * release that reference.
+ */
+ vpending->listener = sk;
+ sock_hold(sk);
+ sock_hold(pending);
+ INIT_DELAYED_WORK(&vpending->dwork, vsock_vmci_pending_work);
+ schedule_delayed_work(&vpending->dwork, HZ);
+
+out:
+ return err;
+}
+
+/*
+ * vsock_vmci_recv_connecting_server --
+ *
+ * Receives packets for sockets in the connecting state on the server side.
+ *
+ * Connecting sockets on the server side can only receive queue pair offer
+ * packets. All others should be treated as cause for closing the connection.
+ *
+ * Note that this assumes the socket lock is held for both sk and pending.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: A queue pair may be created, an attach control packet may be
+ * sent, the socket may transition to the connected state, and a pending caller
+ * in accept() may be woken up.
+ */
+
+static int
+vsock_vmci_recv_connecting_server(struct sock *listener,
+ struct sock *pending, vsock_packet *pkt)
+{
+ vsock_vmci_sock *vpending;
+ struct vmci_handle handle;
+ struct vmci_qp *qpair;
+ bool is_local;
+ u32 flags;
+ vmci_id detach_sub_id;
+ int err;
+ int skerr;
+
+ ASSERT(listener->sk_state == SS_LISTEN);
+ ASSERT(pending->sk_state == SS_CONNECTING);
+
+ vpending = vsock_sk(pending);
+ detach_sub_id = VMCI_INVALID_ID;
+
+ switch (pkt->type) {
+ case VSOCK_PACKET_TYPE_OFFER:
+ if (VMCI_HANDLE_INVALID(pkt->u.handle)) {
+ VSOCK_SEND_RESET(pending, pkt);
+ skerr = EPROTO;
+ err = -EINVAL;
+ goto destroy;
+ }
+ break;
+ default:
+ /* Close and cleanup the connection. */
+ VSOCK_SEND_RESET(pending, pkt);
+ skerr = EPROTO;
+ err = pkt->type == VSOCK_PACKET_TYPE_RST ? 0 : -EINVAL;
+ goto destroy;
+ }
+
+ ASSERT(pkt->type == VSOCK_PACKET_TYPE_OFFER);
+
+ /*
+ * In order to complete the connection we need to attach to the offered
+ * queue pair and send an attach notification. We also subscribe to the
+ * detach event so we know when our peer goes away, and we do that
+ * before attaching so we don't miss an event. If all this succeeds,
+ * we update our state and wakeup anything waiting in accept() for a
+ * connection.
+ */
+
+ /*
+ * We don't care about attach since we ensure the other side has
+ * attached by specifying the ATTACH_ONLY flag below.
+ */
+ err = vmci_event_subscribe(VMCI_EVENT_QP_PEER_DETACH,
+ vsock_vmci_peer_detach_cb,
+ pending, &detach_sub_id);
+ if (err < VMCI_SUCCESS) {
+ VSOCK_SEND_RESET(pending, pkt);
+ err = vsock_vmci_error_to_vsock_error(err);
+ skerr = -err;
+ goto destroy;
+ }
+
+ vpending->detach_sub_id = detach_sub_id;
+
+ /* Now attach to the queue pair the client created. */
+ handle = pkt->u.handle;
+
+ /*
+ * vpending->local_addr always has a context id so we do not need to
+ * worry about VMADDR_CID_ANY in this case.
+ */
+ is_local =
+ vpending->remote_addr.svm_cid == vpending->local_addr.svm_cid;
+ flags = VMCI_QPFLAG_ATTACH_ONLY;
+ flags |= is_local ? VMCI_QPFLAG_LOCAL : 0;
+
+ err = vsock_vmci_queue_pair_alloc(&qpair,
+ &handle,
+ vpending->produce_size,
+ vpending->consume_size,
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->
+ dg.src),
+ flags,
+ vsock_vmci_trusted(
+ vpending,
+ vpending->remote_addr.svm_cid));
+ if (err < 0) {
+ VSOCK_SEND_RESET(pending, pkt);
+ skerr = -err;
+ goto destroy;
+ }
+
+ ASSERT(VMCI_HANDLE_EQUAL(handle, pkt->u.handle));
+ vpending->qp_handle = handle;
+ vpending->qpair = qpair;
+
+ /*
+ * When we send the attach message, we must be ready to handle incoming
+ * control messages on the newly connected socket. So we move the
+ * pending socket to the connected state before sending the attach
+ * message. Otherwise, an incoming packet triggered by the attach being
+ * received by the peer may be processed concurrently with what happens
+ * below after sending the attach message, and that incoming packet
+ * will find the listening socket instead of the (currently) pending
+ * socket. Note that enqueueing the socket increments the reference
+ * count, so even if a reset comes before the connection is accepted,
+ * the socket will be valid until it is removed from the queue.
+ *
+ * If we fail sending the attach below, we remove the socket from the
+ * connected list and move the socket to SS_UNCONNECTED before
+ * releasing the lock, so a pending slow path processing of an incoming
+ * packet will not see the socket in the connected state in that case.
+ */
+ pending->sk_state = SS_CONNECTED;
+
+ vsock_vmci_insert_connected(vsock_connected_sockets_vsk(vpending),
+ pending);
+
+ /* Notify our peer of our attach. */
+ err = VSOCK_SEND_ATTACH(pending, handle);
+ if (err < 0) {
+ vsock_vmci_remove_connected(pending);
+ pr_err("Could not send attach\n");
+ VSOCK_SEND_RESET(pending, pkt);
+ err = vsock_vmci_error_to_vsock_error(err);
+ skerr = -err;
+ goto destroy;
+ }
+
+ /*
+ * We have a connection. Move the now connected socket from the
+ * listener's pending list to the accept queue so callers of accept()
+ * can find it.
+ */
+ vsock_vmci_remove_pending(listener, pending);
+ vsock_vmci_enqueue_accept(listener, pending);
+
+ /*
+ * Callers of accept() will be be waiting on the listening socket, not
+ * the pending socket.
+ */
+ listener->sk_state_change(listener);
+
+ return 0;
+
+destroy:
+ pending->sk_err = skerr;
+ pending->sk_state = SS_UNCONNECTED;
+ /*
+ * As long as we drop our reference, all necessary cleanup will handle
+ * when the cleanup function drops its reference and our destruct
+ * implementation is called. Note that since the listen handler will
+ * remove pending from the pending list upon our failure, the cleanup
+ * function won't drop the additional reference, which is why we do it
+ * here.
+ */
+ sock_put(pending);
+
+ return err;
+}
+
+/*
+ * vsock_vmci_recv_connecting_client --
+ *
+ * Receives packets for sockets in the connecting state on the client side.
+ *
+ * Connecting sockets on the client side should only receive attach packets.
+ * All others should be treated as cause for closing the connection.
+ *
+ * Note that this assumes the socket lock is held for both sk and pending.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: The socket may transition to the connected state and wakeup
+ * the pending caller of connect().
+ */
+
+static int
+vsock_vmci_recv_connecting_client(struct sock *sk, vsock_packet *pkt)
+{
+ vsock_vmci_sock *vsk;
+ int err;
+ int skerr;
+
+ ASSERT(sk->sk_state == SS_CONNECTING);
+
+ vsk = vsock_sk(sk);
+
+ switch (pkt->type) {
+ case VSOCK_PACKET_TYPE_ATTACH:
+ if (VMCI_HANDLE_INVALID(pkt->u.handle) ||
+ !VMCI_HANDLE_EQUAL(pkt->u.handle, vsk->qp_handle)) {
+ skerr = EPROTO;
+ err = -EINVAL;
+ goto destroy;
+ }
+
+ /*
+ * Signify the socket is connected and wakeup the waiter in
+ * connect(). Also place the socket in the connected table for
+ * accounting (it can already be found since it's in the bound
+ * table).
+ */
+ sk->sk_state = SS_CONNECTED;
+ sk->sk_socket->state = SS_CONNECTED;
+ vsock_vmci_insert_connected(vsock_connected_sockets_vsk(vsk),
+ sk);
+ sk->sk_state_change(sk);
+
+ break;
+ case VSOCK_PACKET_TYPE_NEGOTIATE:
+ case VSOCK_PACKET_TYPE_NEGOTIATE2:
+ if (pkt->u.size == 0 ||
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src) !=
+ vsk->remote_addr.svm_cid
+ || pkt->src_port != vsk->remote_addr.svm_port
+ || !VMCI_HANDLE_INVALID(vsk->qp_handle) || vsk->qpair
+ || vsk->produce_size != 0 || vsk->consume_size != 0
+ || vsk->attach_sub_id != VMCI_INVALID_ID
+ || vsk->detach_sub_id != VMCI_INVALID_ID) {
+ skerr = EPROTO;
+ err = -EINVAL;
+
+ goto destroy;
+ }
+
+ err = vsock_vmci_recv_connecting_client_negotiate(sk, pkt);
+ if (err) {
+ skerr = -err;
+ goto destroy;
+ }
+
+ break;
+ case VSOCK_PACKET_TYPE_INVALID:
+ err = vsock_vmci_recv_connecting_client_invalid(sk, pkt);
+ if (err) {
+ skerr = -err;
+ goto destroy;
+ }
+
+ break;
+ case VSOCK_PACKET_TYPE_RST:
+ /*
+ * Older versions of the linux code (WS 6.5 / ESX 4.0) used to
+ * continue processing here after they sent an INVALID packet.
+ * This meant that we got a RST after the INVALID. We ignore a
+ * RST after an INVALID. The common code doesn't send the RST
+ * ... so we can hang if an old version of the common code
+ * fails between getting a REQUEST and sending an OFFER back.
+ * Not much we can do about it... except hope that it doesn't
+ * happen.
+ */
+ if (vsk->ignore_connecting_rst) {
+ vsk->ignore_connecting_rst = false;
+ } else {
+ skerr = ECONNRESET;
+ err = 0;
+ goto destroy;
+ }
+
+ break;
+ default:
+ /* Close and cleanup the connection. */
+ skerr = EPROTO;
+ err = -EINVAL;
+ goto destroy;
+ }
+
+ ASSERT(pkt->type == VSOCK_PACKET_TYPE_ATTACH ||
+ pkt->type == VSOCK_PACKET_TYPE_NEGOTIATE ||
+ pkt->type == VSOCK_PACKET_TYPE_NEGOTIATE2 ||
+ pkt->type == VSOCK_PACKET_TYPE_INVALID ||
+ pkt->type == VSOCK_PACKET_TYPE_RST);
+
+ return 0;
+
+destroy:
+ VSOCK_SEND_RESET(sk, pkt);
+
+ sk->sk_state = SS_UNCONNECTED;
+ sk->sk_err = skerr;
+ sk->sk_error_report(sk);
+ return err;
+}
+
+/*
+ * vsock_vmci_recv_connecting_client_negotiate --
+ *
+ * Handles a negotiate packet for a client in the connecting state.
+ *
+ * Note that this assumes the socket lock is held for both sk and pending.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: The socket may transition to the connected state and wakeup
+ * the pending caller of connect().
+ */
+
+static int
+vsock_vmci_recv_connecting_client_negotiate(struct sock *sk, vsock_packet *pkt)
+{
+ int err;
+ vsock_vmci_sock *vsk;
+ struct vmci_handle handle;
+ struct vmci_qp *qpair;
+ vmci_id attach_sub_id;
+ vmci_id detach_sub_id;
+ bool is_local;
+ u32 flags;
+ bool old_proto = true;
+ bool old_pkt_proto;
+ vsock_proto_version version;
+
+ vsk = vsock_sk(sk);
+ handle = VMCI_INVALID_HANDLE;
+ attach_sub_id = VMCI_INVALID_ID;
+ detach_sub_id = VMCI_INVALID_ID;
+
+ ASSERT(pkt->u.size > 0);
+ ASSERT(vsk->remote_addr.svm_cid ==
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src));
+ ASSERT(vsk->remote_addr.svm_port == pkt->src_port);
+ ASSERT(VMCI_HANDLE_INVALID(vsk->qp_handle));
+ ASSERT(vsk->qpair == NULL);
+ ASSERT(vsk->produce_size == 0);
+ ASSERT(vsk->consume_size == 0);
+ ASSERT(vsk->attach_sub_id == VMCI_INVALID_ID);
+ ASSERT(vsk->detach_sub_id == VMCI_INVALID_ID);
+
+ /*
+ * If we have gotten here then we should be past the point where old
+ * linux vsock could have sent the bogus rst.
+ */
+ vsk->sent_request = false;
+ vsk->ignore_connecting_rst = false;
+
+ /* Verify that we're OK with the proposed queue pair size */
+ if (pkt->u.size < vsk->queue_pair_min_size ||
+ pkt->u.size > vsk->queue_pair_max_size) {
+ err = -EINVAL;
+ goto destroy;
+ }
+
+ /*
+ * At this point we know the CID the peer is using to talk to us.
+ */
+
+ if (vsk->local_addr.svm_cid == VMADDR_CID_ANY)
+ vsk->local_addr.svm_cid =
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.dst);
+
+ /*
+ * Setup the notify ops to be the highest supported version that both
+ * the server and the client support.
+ */
+
+ if (vsock_vmci_old_proto_override(&old_pkt_proto)) {
+ old_proto = old_pkt_proto;
+ } else {
+ if (pkt->type == VSOCK_PACKET_TYPE_NEGOTIATE)
+ old_proto = true;
+ else if (pkt->type == VSOCK_PACKET_TYPE_NEGOTIATE2)
+ old_proto = false;
+
+ }
+
+ if (old_proto)
+ version = VSOCK_PROTO_INVALID;
+ else
+ version = pkt->proto;
+
+ if (!vsock_vmci_proto_to_notify_struct(sk, &version, old_proto)) {
+ err = -EINVAL;
+ goto destroy;
+ }
+
+ /*
+ * Subscribe to attach and detach events first.
+ *
+ * XXX We attach once for each queue pair created for now so it is easy
+ * to find the socket (it's provided), but later we should only
+ * subscribe once and add a way to lookup sockets by queue pair handle.
+ */
+ err = vmci_event_subscribe(VMCI_EVENT_QP_PEER_ATTACH,
+ vsock_vmci_peer_attach_cb,
+ sk, &attach_sub_id);
+ if (err < VMCI_SUCCESS) {
+ err = vsock_vmci_error_to_vsock_error(err);
+ goto destroy;
+ }
+
+ err = vmci_event_subscribe(VMCI_EVENT_QP_PEER_DETACH,
+ vsock_vmci_peer_detach_cb,
+ sk, &detach_sub_id);
+ if (err < VMCI_SUCCESS) {
+ err = vsock_vmci_error_to_vsock_error(err);
+ goto destroy;
+ }
+
+ /* Make VMCI select the handle for us. */
+ handle = VMCI_INVALID_HANDLE;
+ is_local = vsk->remote_addr.svm_cid == vsk->local_addr.svm_cid;
+ flags = is_local ? VMCI_QPFLAG_LOCAL : 0;
+
+ err = vsock_vmci_queue_pair_alloc(&qpair,
+ &handle,
+ pkt->u.size,
+ pkt->u.size,
+ vsk->remote_addr.svm_cid,
+ flags,
+ vsock_vmci_trusted(
+ vsk,
+ vsk->
+ remote_addr.svm_cid));
+ if (err < 0)
+ goto destroy;
+
+ err = VSOCK_SEND_QP_OFFER(sk, handle);
+ if (err < 0) {
+ err = vsock_vmci_error_to_vsock_error(err);
+ goto destroy;
+ }
+
+ vsk->qp_handle = handle;
+ vsk->qpair = qpair;
+
+ vsk->produce_size = vsk->consume_size = pkt->u.size;
+
+ vsk->attach_sub_id = attach_sub_id;
+ vsk->detach_sub_id = detach_sub_id;
+
+ NOTIFYCALL(vsk, process_negotiate, sk);
+
+ return 0;
+
+destroy:
+ if (attach_sub_id != VMCI_INVALID_ID) {
+ vmci_event_unsubscribe(attach_sub_id);
+ ASSERT(vsk->attach_sub_id == VMCI_INVALID_ID);
+ }
+
+ if (detach_sub_id != VMCI_INVALID_ID) {
+ vmci_event_unsubscribe(detach_sub_id);
+ ASSERT(vsk->detach_sub_id == VMCI_INVALID_ID);
+ }
+
+ if (!VMCI_HANDLE_INVALID(handle)) {
+ vmci_qpair_detach(&qpair);
+ ASSERT(VMCI_HANDLE_INVALID(vsk->qp_handle));
+ }
+
+ return err;
+}
+
+/*
+ * vsock_vmci_recv_connecting_client_invalid --
+ *
+ * Handles an invalid packet for a client in the connecting state.
+ *
+ * Note that this assumes the socket lock is held for both sk and pending.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_recv_connecting_client_invalid(struct sock *sk, vsock_packet *pkt)
+{
+ int err = 0;
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ if (vsk->sent_request) {
+ vsk->sent_request = false;
+ vsk->ignore_connecting_rst = true;
+
+ err = VSOCK_SEND_CONN_REQUEST(sk, vsk->queue_pair_size);
+ if (err < 0)
+ err = vsock_vmci_error_to_vsock_error(err);
+ else
+ err = 0;
+
+ }
+
+ return err;
+}
+
+/*
+ * vsock_vmci_recv_connected --
+ *
+ * Receives packets for sockets in the connected state.
+ *
+ * Connected sockets should only ever receive detach, wrote, read, or reset
+ * control messages. Others are treated as errors that are ignored.
+ *
+ * Wrote and read signify that the peer has produced or consumed, respectively.
+ *
+ * Detach messages signify that the connection is being closed cleanly and
+ * reset messages signify that the connection is being closed in error.
+ *
+ * Note that this assumes the socket lock is held.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: A queue pair may be created, an offer control packet sent, and
+ * the socket may transition to the connecting state.
+ *
+ */
+
+static int vsock_vmci_recv_connected(struct sock *sk, vsock_packet *pkt)
+{
+ vsock_vmci_sock *vsk;
+ bool pkt_processed = false;
+
+ ASSERT(sk->sk_state == SS_CONNECTED);
+
+ /*
+ * In cases where we are closing the connection, it's sufficient to
+ * mark the state change (and maybe error) and wake up any waiting
+ * threads. Since this is a connected socket, it's owned by a user
+ * process and will be cleaned up when the failure is passed back on
+ * the current or next system call. Our system call implementations
+ * must therefore check for error and state changes on entry and when
+ * being awoken.
+ */
+ switch (pkt->type) {
+ case VSOCK_PACKET_TYPE_SHUTDOWN:
+ if (pkt->u.mode) {
+ vsk = vsock_sk(sk);
+
+ vsk->peer_shutdown |= pkt->u.mode;
+ sk->sk_state_change(sk);
+ }
+ break;
+
+ case VSOCK_PACKET_TYPE_RST:
+ vsk = vsock_sk(sk);
+ /*
+ * It is possible that we sent our peer a message (e.g a
+ * WAITING_READ) right before we got notified that the peer had
+ * detached. If that happens then we can get a RST pkt back
+ * from our peer even though there is data available for us to
+ * read. In that case, don't shutdown the socket completely but
+ * instead allow the local client to finish reading data off
+ * the queuepair. Always treat a RST pkt in connected mode like
+ * a clean shutdown.
+ */
+ sock_set_flag(sk, SOCK_DONE);
+ vsk->peer_shutdown = SHUTDOWN_MASK;
+ if (vsock_vmci_stream_has_data(vsk) <= 0)
+ sk->sk_state = SS_DISCONNECTING;
+
+ sk->sk_state_change(sk);
+ break;
+
+ default:
+ vsk = vsock_sk(sk);
+ NOTIFYCALL(vsk, handle_notify_pkt, sk, pkt, false, NULL, NULL,
+ &pkt_processed);
+ if (!pkt_processed)
+ return -EINVAL;
+
+ break;
+ }
+
+ return 0;
+}
+
+/*
+ * __vsock_vmci_send_control_pkt --
+ *
+ * Common code to send a control packet.
+ *
+ * Results: Size of datagram sent on success, negative error code otherwise. If
+ * convert_error is true, error code is a vsock error, otherwise, result is a
+ * VMCI error code.
+ *
+ * Side effects: None.
+ */
+
+static int
+__vsock_vmci_send_control_pkt(vsock_packet *pkt,
+ struct sockaddr_vm *src,
+ struct sockaddr_vm *dst,
+ vsock_packet_type type,
+ u64 size,
+ u64 mode,
+ vsock_waiting_info *wait,
+ vsock_proto_version proto,
+ struct vmci_handle handle, bool convert_error)
+{
+ int err;
+
+ BUG_ON(vsock_addr_validate(src) != 0);
+ BUG_ON(vsock_addr_validate(dst) != 0);
+
+ vsock_packet_init(pkt, src, dst, type, size, mode, wait, proto, handle);
+ LOG_PACKET(pkt);
+ VSOCK_STATS_CTLPKT_LOG(pkt->type);
+ err = vmci_datagram_send(&pkt->dg);
+ if (convert_error && (err < 0))
+ return vsock_vmci_error_to_vsock_error(err);
+
+ return err;
+}
+
+/*
+ * vsock_vmci_reply_control_pkt_fast --
+ *
+ * Sends a control packet back to the source of an incoming packet. The control
+ * packet is allocated in the stack.
+ *
+ * Results: Size of datagram sent on success, negative error code otherwise.
+ *
+ * Side effects: None.
+ */
+
+int
+vsock_vmci_reply_control_pkt_fast(vsock_packet *pkt,
+ vsock_packet_type type,
+ u64 size,
+ u64 mode,
+ vsock_waiting_info *wait,
+ struct vmci_handle handle)
+{
+ vsock_packet reply;
+ struct sockaddr_vm src, dst;
+
+ if (pkt->type == VSOCK_PACKET_TYPE_RST) {
+ return 0;
+ } else {
+ vsock_packet_get_addresses(pkt, &src, &dst);
+ return __vsock_vmci_send_control_pkt(&reply, &src, &dst, type,
+ size, mode, wait,
+ VSOCK_PROTO_INVALID,
+ handle, true);
+ }
+}
+
+/*
+ * vsock_vmci_send_control_pkt_bh --
+ *
+ * Sends a control packet from bottom-half context. The control packet is
+ * static data to minimize the resource cost.
+ *
+ * Results: Size of datagram sent on success, negative error code otherwise.
+ * Note that we return a VMCI error message since that's what callers will need
+ * to provide.
+ *
+ * Side effects: None.
+ */
+
+int
+vsock_vmci_send_control_pkt_bh(struct sockaddr_vm *src,
+ struct sockaddr_vm *dst,
+ vsock_packet_type type,
+ u64 size,
+ u64 mode,
+ vsock_waiting_info *wait,
+ struct vmci_handle handle)
+{
+ /*
+ * Note that it is safe to use a single packet across all CPUs since
+ * two tasklets of the same type are guaranteed to not ever run
+ * simultaneously. If that ever changes, or VMCI stops using tasklets,
+ * we can use per-cpu packets.
+ */
+ static vsock_packet pkt;
+
+ return __vsock_vmci_send_control_pkt(&pkt, src, dst, type,
+ size, mode, wait,
+ VSOCK_PROTO_INVALID, handle,
+ false);
+}
+
+/*
+ * vsock_vmci_send_control_pkt --
+ *
+ * Sends a control packet.
+ *
+ * Results: Size of datagram sent on success, negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+int
+vsock_vmci_send_control_pkt(struct sock *sk,
+ vsock_packet_type type,
+ u64 size,
+ u64 mode,
+ vsock_waiting_info *wait,
+ vsock_proto_version proto,
+ struct vmci_handle handle)
+{
+ vsock_packet *pkt;
+ vsock_vmci_sock *vsk;
+ int err;
+
+ vsk = vsock_sk(sk);
+
+ if (!vsock_addr_bound(&vsk->local_addr))
+ return -EINVAL;
+
+ if (!vsock_addr_bound(&vsk->remote_addr))
+ return -EINVAL;
+
+ pkt = kmalloc(sizeof *pkt, GFP_KERNEL);
+ if (!pkt)
+ return -ENOMEM;
+
+ err =
+ __vsock_vmci_send_control_pkt(pkt, &vsk->local_addr,
+ &vsk->remote_addr, type, size, mode,
+ wait, proto, handle, true);
+ kfree(pkt);
+
+ return err;
+}
+
+/*
+ * __vsock_vmci_bind --
+ *
+ * Common functionality needed to bind the specified address to the VSocket.
+ * If VMADDR_CID_ANY or VMADDR_PORT_ANY are specified, the context ID or port
+ * are selected automatically.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: On success, a new datagram handle is created.
+ */
+
+static int __vsock_vmci_bind(struct sock *sk, struct sockaddr_vm *addr)
+{
+ static unsigned int port = LAST_RESERVED_PORT + 1;
+ struct sockaddr_vm new_addr;
+ vsock_vmci_sock *vsk;
+ vmci_id cid;
+ int err;
+
+ vsk = vsock_sk(sk);
+
+ /* First ensure this socket isn't already bound. */
+ if (vsock_addr_bound(&vsk->local_addr))
+ return -EINVAL;
+
+ /*
+ * Now bind to the provided address or select appropriate values if
+ * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
+ * like AF_INET prevents binding to a non-local IP address (in most
+ * cases), we only allow binding to the local CID.
+ */
+ vsock_addr_init(&new_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+
+ cid = vmci_get_context_id();
+ if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
+ return -EADDRNOTAVAIL;
+
+ new_addr.svm_cid = addr->svm_cid;
+
+ switch (sk->sk_socket->type) {
+ case SOCK_STREAM:
+ spin_lock_bh(&vsock_table_lock);
+
+ if (addr->svm_port == VMADDR_PORT_ANY) {
+ bool found = false;
+ unsigned int i;
+
+ for (i = 0; i < MAX_PORT_RETRIES; i++) {
+ if (port <= LAST_RESERVED_PORT)
+ port = LAST_RESERVED_PORT + 1;
+
+ new_addr.svm_port = port++;
+
+ if (!__vsock_vmci_find_bound_socket
+ (&new_addr)) {
+ found = true;
+ break;
+ }
+ }
+
+ if (!found) {
+ err = -EADDRNOTAVAIL;
+ goto out;
+ }
+ } else {
+ /*
+ * If port is in reserved range, ensure caller
+ * has necessary privileges.
+ */
+ if (addr->svm_port <= LAST_RESERVED_PORT &&
+ !capable(CAP_NET_BIND_SERVICE)) {
+ err = -EACCES;
+ goto out;
+ }
+
+ new_addr.svm_port = addr->svm_port;
+ if (__vsock_vmci_find_bound_socket(&new_addr)) {
+ err = -EADDRINUSE;
+ goto out;
+ }
+
+ }
+ break;
+ case SOCK_DGRAM: {
+ u32 flags = 0;
+
+ /*
+ * VMCI will select a resource ID for us if we provide
+ * VMCI_INVALID_ID.i
+ */
+ new_addr.svm_port = addr->svm_port == VMADDR_PORT_ANY ?
+ VMCI_INVALID_ID : addr->svm_port;
+
+ if (new_addr.svm_port <= LAST_RESERVED_PORT &&
+ !capable(CAP_NET_BIND_SERVICE)) {
+ err = -EACCES;
+ goto out;
+ }
+
+ if (new_addr.svm_cid == VMADDR_CID_ANY)
+ flags = VMCI_FLAG_ANYCID_DG_HND;
+
+ err = vsock_vmci_datagram_create_hnd(new_addr.svm_port,
+ flags,
+ vsock_vmci_recv_dgram_cb,
+ sk, &vsk->dg_handle);
+ if (err < VMCI_SUCCESS) {
+ err = vsock_vmci_error_to_vsock_error(err);
+ goto out;
+ }
+
+ new_addr.svm_port = VMCI_HANDLE_TO_RESOURCE_ID(vsk->dg_handle);
+ break;
+ }
+ default:
+ err = -EINVAL;
+ goto out;
+ }
+
+ vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
+
+ /*
+ * Remove stream sockets from the unbound list and add them to the hash
+ * table for easy lookup by its address. The unbound list is simply an
+ * extra entry at the end of the hash table, a trick used by AF_UNIX.
+ */
+ if (sk->sk_socket->type == SOCK_STREAM) {
+ __vsock_vmci_remove_bound(sk);
+ __vsock_vmci_insert_bound(vsock_bound_sockets(&vsk->local_addr),
+ sk);
+ spin_unlock_bh(&vsock_table_lock);
+ }
+
+ BUG_ON(vsock_addr_validate(&vsk->local_addr) != 0);
+ return 0;
+
+out:
+ if (sk->sk_socket->type == SOCK_STREAM)
+ spin_unlock_bh(&vsock_table_lock);
+
+ return err;
+}
+
+/*
+ * __vsock_vmci_create --
+ *
+ * Does the work to create the sock structure. Note: If sock is NULL then the
+ * type field must be non-zero. Otherwise, sock is non-NULL and the type of
+ * sock is used in the newly created socket.
+ *
+ * Results: sock structure on success, NULL on failure.
+ *
+ * Side effects: Allocated sk is added to the unbound sockets list iff it is
+ * owned by a struct socket.
+ */
+
+static struct sock *__vsock_vmci_create(struct net *net,
+ struct socket *sock,
+ struct sock *parent,
+ gfp_t priority, unsigned short type)
+{
+ struct sock *sk;
+ vsock_vmci_sock *psk;
+ vsock_vmci_sock *vsk;
+
+ ASSERT((sock && !type) || (!sock && type));
+
+ vsk = NULL;
+
+ /*
+ * From 2.6.9 to until 2.6.12 sk_alloc() used a cache in the protocol
+ * structure, but you still had to specify the size and cache yourself.
+ * Most recently (in 2.6.24), sk_alloc() was changed to expect the
+ * network namespace, and the option to zero the sock was dropped.
+ */
+ sk = sk_alloc(net, vsock_vmci_family_ops.family, priority,
+ &vsock_vmci_proto);
+ if (!sk)
+ return NULL;
+
+ sock_init_data(sock, sk);
+
+ /*
+ * sk->sk_type is normally set in sock_init_data, but only if sock is
+ * non-NULL. We make sure that our sockets always have a type by
+ * setting it here if needed.
+ */
+ if (!sock)
+ sk->sk_type = type;
+
+ vsk = vsock_sk(sk);
+ vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+ vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+
+ sk->sk_destruct = vsock_vmci_sk_destruct;
+ sk->sk_backlog_rcv = vsock_vmci_queue_rcv_skb;
+ sk->sk_state = 0;
+ sock_reset_flag(sk, SOCK_DONE);
+
+ INIT_LIST_HEAD(&vsk->bound_table);
+ INIT_LIST_HEAD(&vsk->connected_table);
+ vsk->dg_handle = VMCI_INVALID_HANDLE;
+ vsk->qp_handle = VMCI_INVALID_HANDLE;
+ vsk->qpair = NULL;
+ vsk->produce_size = vsk->consume_size = 0;
+ vsk->listener = NULL;
+ INIT_LIST_HEAD(&vsk->pending_links);
+ INIT_LIST_HEAD(&vsk->accept_queue);
+ vsk->rejected = false;
+ vsk->sent_request = false;
+ vsk->ignore_connecting_rst = false;
+ vsk->attach_sub_id = vsk->detach_sub_id = VMCI_INVALID_ID;
+ vsk->peer_shutdown = 0;
+
+ if (parent) {
+ psk = vsock_sk(parent);
+ vsk->trusted = psk->trusted;
+ vsk->owner = get_cred(psk->owner);
+ vsk->queue_pair_size = psk->queue_pair_size;
+ vsk->queue_pair_min_size = psk->queue_pair_min_size;
+ vsk->queue_pair_max_size = psk->queue_pair_max_size;
+ vsk->connect_timeout = psk->connect_timeout;
+ } else {
+ vsk->trusted = capable(CAP_NET_ADMIN);
+ vsk->owner = get_current_cred();
+ vsk->queue_pair_size = VSOCK_DEFAULT_QP_SIZE;
+ vsk->queue_pair_min_size = VSOCK_DEFAULT_QP_SIZE_MIN;
+ vsk->queue_pair_max_size = VSOCK_DEFAULT_QP_SIZE_MAX;
+ vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
+ }
+
+ vsk->notify_ops = NULL;
+
+ if (sock)
+ vsock_vmci_insert_bound(vsock_unbound_sockets, sk);
+
+ return sk;
+}
+
+/*
+ * __vsock_vmci_release --
+ *
+ * Releases the provided socket.
+ *
+ * Results: None.
+ *
+ * Side effects: Any pending sockets are also released.
+ */
+
+static void __vsock_vmci_release(struct sock *sk)
+{
+ if (sk) {
+ struct sk_buff *skb;
+ struct sock *pending;
+ struct vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+ pending = NULL; /* Compiler warning. */
+
+ if (vsock_vmci_in_bound_table(sk))
+ vsock_vmci_remove_bound(sk);
+
+ if (vsock_vmci_in_connected_table(sk))
+ vsock_vmci_remove_connected(sk);
+
+ if (!VMCI_HANDLE_INVALID(vsk->dg_handle)) {
+ vmci_datagram_destroy_handle(vsk->dg_handle);
+ vsk->dg_handle = VMCI_INVALID_HANDLE;
+ }
+
+ lock_sock(sk);
+ sock_orphan(sk);
+ sk->sk_shutdown = SHUTDOWN_MASK;
+
+ while ((skb = skb_dequeue(&sk->sk_receive_queue)))
+ kfree_skb(skb);
+
+ /* Clean up any sockets that never were accepted. */
+ while ((pending = vsock_vmci_dequeue_accept(sk)) != NULL) {
+ __vsock_vmci_release(pending);
+ sock_put(pending);
+ }
+
+ release_sock(sk);
+ sock_put(sk);
+ }
+}
+
+/*
+ * Sock operations.
+ */
+
+/*
+ * vsock_vmci_sk_destruct --
+ *
+ * Destroys the provided socket. This is called by sk_free(), which is invoke
+ * when the reference count of the socket drops to zero.
+ *
+ * Results: None.
+ *
+ * Side effects: Socket count is decremented.
+ */
+
+static void vsock_vmci_sk_destruct(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ if (vsk->attach_sub_id != VMCI_INVALID_ID) {
+ vmci_event_unsubscribe(vsk->attach_sub_id);
+ vsk->attach_sub_id = VMCI_INVALID_ID;
+ }
+
+ if (vsk->detach_sub_id != VMCI_INVALID_ID) {
+ vmci_event_unsubscribe(vsk->detach_sub_id);
+ vsk->detach_sub_id = VMCI_INVALID_ID;
+ }
+
+ if (!VMCI_HANDLE_INVALID(vsk->qp_handle)) {
+ vmci_qpair_detach(&vsk->qpair);
+ vsk->qp_handle = VMCI_INVALID_HANDLE;
+ vsk->produce_size = vsk->consume_size = 0;
+ }
+
+ /*
+ * Each list entry holds a reference on the socket, so we should not
+ * even be here if the socket is in one of our lists. If we are we
+ * have a stray sock_put() that needs to go away.
+ */
+ ASSERT(!vsock_vmci_in_bound_table(sk));
+ ASSERT(!vsock_vmci_in_connected_table(sk));
+ ASSERT(!vsock_vmci_is_pending(sk));
+ ASSERT(!vsock_vmci_in_accept_queue(sk));
+
+ /*
+ * When clearing these addresses, there's no need to set the family and
+ * possibly register the address family with the kernel.
+ */
+ vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+ vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+
+ NOTIFYCALL(vsk, socket_destruct, sk);
+
+ put_cred(vsk->owner);
+
+ VSOCK_STATS_CTLPKT_DUMP_ALL();
+ VSOCK_STATS_HIST_DUMP_ALL();
+ VSOCK_STATS_TOTALS_DUMP_ALL();
+}
+
+/*
+ * vsock_vmci_queue_rcv_skb --
+ *
+ * Receives skb on the socket's receive queue.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
+{
+ int err;
+
+ err = sock_queue_rcv_skb(sk, skb);
+ if (err)
+ kfree_skb(skb);
+
+ return err;
+}
+
+/*
+ * vsock_vmci_register_with_vmci --
+ *
+ * Registers with the VMCI device, and creates control message and event
+ * handlers.
+ *
+ * Results: Zero on success, error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_register_with_vmci(void)
+{
+ int err = 0;
+
+ /*
+ * Create the datagram handle that we will use to send and receive all
+ * VSocket control messages for this context.
+ */
+ err = vsock_vmci_datagram_create_hnd(VSOCK_PACKET_RID,
+ VMCI_FLAG_ANYCID_DG_HND,
+ vsock_vmci_recv_stream_cb, NULL,
+ &vmci_stream_handle);
+ if (err < VMCI_SUCCESS) {
+ pr_err("Unable to create datagram handle. (%d)\n",
+ err);
+ err = vsock_vmci_error_to_vsock_error(err);
+ goto out;
+ }
+
+ err = vmci_event_subscribe(VMCI_EVENT_QP_RESUMED,
+ vsock_vmci_qp_resumed_cb,
+ NULL, &qp_resumed_sub_id);
+ if (err < VMCI_SUCCESS) {
+ pr_err("Unable to subscribe to resumed event. (%d)\n",
+ err);
+ err = vsock_vmci_error_to_vsock_error(err);
+ qp_resumed_sub_id = VMCI_INVALID_ID;
+ goto out;
+ }
+
+out:
+ if (err != 0)
+ vsock_vmci_unregister_with_vmci();
+
+ return err;
+}
+
+/*
+ * vsock_vmci_unregister_with_vmci --
+ *
+ * Destroys control message and event handlers, and unregisters with the VMCI
+ * device
+ *
+ * Results: None.
+ *
+ * Side effects: Our socket implementation is no longer accessible.
+ */
+
+static void vsock_vmci_unregister_with_vmci(void)
+{
+ if (!VMCI_HANDLE_INVALID(vmci_stream_handle)) {
+ if (vmci_datagram_destroy_handle(vmci_stream_handle) !=
+ VMCI_SUCCESS)
+ pr_err("Couldn't destroy datagram handle.\n");
+
+ vmci_stream_handle = VMCI_INVALID_HANDLE;
+ }
+
+ if (qp_resumed_sub_id != VMCI_INVALID_ID) {
+ vmci_event_unsubscribe(qp_resumed_sub_id);
+ qp_resumed_sub_id = VMCI_INVALID_ID;
+ }
+}
+
+/*
+ * vsock_vmci_stream_has_data --
+ *
+ * Gets the amount of data available for a given stream socket's consume queue.
+ *
+ * Note that this assumes the socket lock is held.
+ *
+ * Results: The amount of data available or a VMCI error code on failure.
+ *
+ * Side effects: None.
+ */
+
+s64 vsock_vmci_stream_has_data(vsock_vmci_sock *vsk)
+{
+ return vmci_qpair_consume_buf_ready(vsk->qpair);
+}
+
+/*
+ * vsock_vmci_stream_has_space --
+ *
+ * Gets the amount of space available for a give stream socket's produce queue.
+ *
+ * Note that this assumes the socket lock is held.
+ *
+ * Results: The amount of space available or a VMCI error code on failure.
+ *
+ * Side effects: None.
+ */
+
+s64 vsock_vmci_stream_has_space(vsock_vmci_sock *vsk)
+{
+ return vmci_qpair_produce_free_space(vsk->qpair);
+}
+
+/*
+ * Socket operations.
+ */
+
+/*
+ *
+ * vsock_vmci_release --
+ *
+ * Releases the provided socket by freeing the contents of its queue. This is
+ * called when a user process calls close(2) on the socket.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_release(struct socket *sock)
+{
+ __vsock_vmci_release(sock->sk);
+ sock->sk = NULL;
+ sock->state = SS_FREE;
+
+ return 0;
+}
+
+/*
+ * vsock_vmci_bind --
+ *
+ * Binds the provided address to the provided socket.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
+{
+ int err;
+ struct sock *sk;
+ struct sockaddr_vm *vmci_addr;
+
+ sk = sock->sk;
+
+ if (vsock_addr_cast(addr, addr_len, &vmci_addr) != 0)
+ return -EINVAL;
+
+ lock_sock(sk);
+ err = __vsock_vmci_bind(sk, vmci_addr);
+ release_sock(sk);
+
+ return err;
+}
+
+/*
+ * vsock_vmci_dgram_connect --
+ *
+ * Connects a datagram socket. This can be called multiple times to change the
+ * socket's association and can be called with a sockaddr whose family is set
+ * to AF_UNSPEC to dissolve any existing association.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_dgram_connect(struct socket *sock,
+ struct sockaddr *addr, int addr_len, int flags)
+{
+ int err;
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+ struct sockaddr_vm *remote_addr;
+
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+
+ err = vsock_addr_cast(addr, addr_len, &remote_addr);
+ if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
+ lock_sock(sk);
+ vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
+ VMADDR_PORT_ANY);
+ sock->state = SS_UNCONNECTED;
+ release_sock(sk);
+ return 0;
+ } else if (err != 0)
+ return -EINVAL;
+
+ lock_sock(sk);
+
+ if (!vsock_addr_bound(&vsk->local_addr)) {
+ struct sockaddr_vm local_addr;
+
+ vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+ err = __vsock_vmci_bind(sk, &local_addr);
+ if (err != 0)
+ goto out;
+
+ }
+
+ if (!vsock_addr_socket_context_dgram(remote_addr->svm_cid,
+ remote_addr->svm_port)) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ memcpy(&vsk->remote_addr, remote_addr, sizeof vsk->remote_addr);
+ sock->state = SS_CONNECTED;
+
+out:
+ release_sock(sk);
+ return err;
+}
+
+/*
+ * vsock_vmci_connect_timeout --
+ *
+ * Asynchronous connection attempts schedule this timeout function to notify
+ * the connector of an unsuccessfull connection attempt. If the socket is still
+ * in the connecting state and hasn't been closed, we mark the socket as timed
+ * out. Otherwise, we do nothing.
+ *
+ * Results: None.
+ *
+ * Side effects: May destroy the socket.
+ */
+
+static void vsock_vmci_connect_timeout(struct work_struct *work)
+{
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+
+ vsk = container_of(work, vsock_vmci_sock, dwork.work);
+ sk = sk_vsock(vsk);
+
+ lock_sock(sk);
+ if (sk->sk_state == SS_CONNECTING &&
+ (sk->sk_shutdown != SHUTDOWN_MASK)) {
+ sk->sk_state = SS_UNCONNECTED;
+ sk->sk_err = ETIMEDOUT;
+ sk->sk_error_report(sk);
+ }
+ release_sock(sk);
+
+ sock_put(sk);
+}
+
+/*
+ * vsock_vmci_stream_connect --
+ *
+ * Connects a stream socket.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_stream_connect(struct socket *sock,
+ struct sockaddr *addr, int addr_len, int flags)
+{
+ int err;
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+ struct sockaddr_vm *remote_addr;
+ long timeout;
+ bool old_pkt_proto = false;
+ DEFINE_WAIT(wait);
+
+ err = 0;
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+
+ lock_sock(sk);
+
+ /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
+ switch (sock->state) {
+ case SS_CONNECTED:
+ err = -EISCONN;
+ goto out;
+ case SS_DISCONNECTING:
+ err = -EINVAL;
+ goto out;
+ case SS_CONNECTING:
+ /*
+ * This continues on so we can move sock into the SS_CONNECTED
+ * state once the connection has completed (at which point err
+ * will be set to zero also). Otherwise, we will either wait
+ * for the connection or return -EALREADY should this be a
+ * non-blocking call.
+ */
+ err = -EALREADY;
+ break;
+ default:
+ ASSERT(sk->sk_state == SS_FREE ||
+ sk->sk_state == SS_UNCONNECTED ||
+ sk->sk_state == SS_LISTEN);
+ if ((sk->sk_state == SS_LISTEN) ||
+ vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ /*
+ * The hypervisor and well-known contexts do not have socket
+ * endpoints.
+ */
+ if (!vsock_addr_socket_context_stream(remote_addr->svm_cid)) {
+ err = -ENETUNREACH;
+ goto out;
+ }
+
+ /* Set the remote address that we are connecting to. */
+ memcpy(&vsk->remote_addr, remote_addr, sizeof vsk->remote_addr);
+
+ /* Autobind this socket to the local address if necessary. */
+ if (!vsock_addr_bound(&vsk->local_addr)) {
+ struct sockaddr_vm local_addr;
+
+ vsock_addr_init(&local_addr, VMADDR_CID_ANY,
+ VMADDR_PORT_ANY);
+ err = __vsock_vmci_bind(sk, &local_addr);
+ if (err != 0)
+ goto out;
+
+ }
+
+ sk->sk_state = SS_CONNECTING;
+
+ if (vsock_vmci_old_proto_override(&old_pkt_proto)
+ && old_pkt_proto) {
+ err = VSOCK_SEND_CONN_REQUEST(sk, vsk->queue_pair_size);
+ if (err < 0) {
+ sk->sk_state = SS_UNCONNECTED;
+ goto out;
+ }
+ } else {
+ int supported_proto_versions =
+ vsock_vmci_new_proto_supported_versions();
+ err =
+ VSOCK_SEND_CONN_REQUEST2(sk, vsk->queue_pair_size,
+ supported_proto_versions);
+ if (err < 0) {
+ sk->sk_state = SS_UNCONNECTED;
+ goto out;
+ }
+
+ vsk->sent_request = true;
+ }
+
+ /*
+ * Mark sock as connecting and set the error code to in
+ * progress in case this is a non-blocking connect.
+ */
+ sock->state = SS_CONNECTING;
+ err = -EINPROGRESS;
+ }
+
+ /*
+ * The receive path will handle all communication until we are able to
+ * enter the connected state. Here we wait for the connection to be
+ * completed or a notification of an error.
+ */
+ timeout = vsk->connect_timeout;
+ prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
+
+ while (sk->sk_state != SS_CONNECTED && sk->sk_err == 0) {
+ if (flags & O_NONBLOCK) {
+ /*
+ * If we're not going to block, we schedule a timeout
+ * function to generate a timeout on the connection
+ * attempt, in case the peer doesn't respond in a
+ * timely manner. We hold on to the socket until the
+ * timeout fires.
+ */
+ sock_hold(sk);
+ INIT_DELAYED_WORK(&vsk->dwork,
+ vsock_vmci_connect_timeout);
+ schedule_delayed_work(&vsk->dwork, timeout);
+
+ /* Skip ahead to preserve error code set above. */
+ goto out_wait;
+ }
+
+ release_sock(sk);
+ timeout = schedule_timeout(timeout);
+ lock_sock(sk);
+
+ if (signal_pending(current)) {
+ err = sock_intr_errno(timeout);
+ goto out_wait_error;
+ } else if (timeout == 0) {
+ err = -ETIMEDOUT;
+ goto out_wait_error;
+ }
+
+ prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
+ }
+
+ if (sk->sk_err) {
+ err = -sk->sk_err;
+ goto out_wait_error;
+ } else {
+ ASSERT(sk->sk_state == SS_CONNECTED);
+ err = 0;
+ }
+
+out_wait:
+ finish_wait(sk_sleep(sk), &wait);
+out:
+ release_sock(sk);
+ return err;
+
+out_wait_error:
+ sk->sk_state = SS_UNCONNECTED;
+ sock->state = SS_UNCONNECTED;
+ goto out_wait;
+}
+
+/*
+ * vsock_vmci_accept --
+ *
+ * Accepts next available connection request for this socket.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_accept(struct socket *sock, struct socket *newsock, int flags)
+{
+ struct sock *listener;
+ int err;
+ struct sock *connected;
+ vsock_vmci_sock *vconnected;
+ long timeout;
+ DEFINE_WAIT(wait);
+
+ err = 0;
+ listener = sock->sk;
+
+ lock_sock(listener);
+
+ if (sock->type != SOCK_STREAM) {
+ err = -EOPNOTSUPP;
+ goto out;
+ }
+
+ if (listener->sk_state != SS_LISTEN) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ /*
+ * Wait for children sockets to appear; these are the new sockets
+ * created upon connection establishment.
+ */
+ timeout = sock_sndtimeo(listener, flags & O_NONBLOCK);
+ prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
+
+ while ((connected = vsock_vmci_dequeue_accept(listener)) == NULL &&
+ listener->sk_err == 0) {
+ release_sock(listener);
+ timeout = schedule_timeout(timeout);
+ lock_sock(listener);
+
+ if (signal_pending(current)) {
+ err = sock_intr_errno(timeout);
+ goto out_wait;
+ } else if (timeout == 0) {
+ err = -EAGAIN;
+ goto out_wait;
+ }
+
+ prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
+ }
+
+ if (listener->sk_err)
+ err = -listener->sk_err;
+
+ if (connected) {
+ listener->sk_ack_backlog--;
+
+ lock_sock(connected);
+ vconnected = vsock_sk(connected);
+
+ /*
+ * If the listener socket has received an error, then we should
+ * reject this socket and return. Note that we simply mark the
+ * socket rejected, drop our reference, and let the cleanup
+ * function handle the cleanup; the fact that we found it in
+ * the listener's accept queue guarantees that the cleanup
+ * function hasn't run yet.
+ */
+ if (err) {
+ vconnected->rejected = true;
+ release_sock(connected);
+ sock_put(connected);
+ goto out_wait;
+ }
+
+ newsock->state = SS_CONNECTED;
+ sock_graft(connected, newsock);
+ release_sock(connected);
+ sock_put(connected);
+ }
+
+out_wait:
+ finish_wait(sk_sleep(listener), &wait);
+out:
+ release_sock(listener);
+ return err;
+}
+
+/*
+ * vsock_vmci_getname --
+ *
+ * Provides the local or remote address for the socket.
+ *
+ * Results: Zero on success, negative error code otherwise.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_getname(struct socket *sock,
+ struct sockaddr *addr, int *addr_len, int peer)
+{
+ int err;
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+ struct sockaddr_vm *vmci_addr;
+
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+ err = 0;
+
+ lock_sock(sk);
+
+ if (peer) {
+ if (sock->state != SS_CONNECTED) {
+ err = -ENOTCONN;
+ goto out;
+ }
+ vmci_addr = &vsk->remote_addr;
+ } else {
+ vmci_addr = &vsk->local_addr;
+ }
+
+ if (!vmci_addr) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ /*
+ * sys_getsockname() and sys_getpeername() pass us a
+ * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
+ * that macro is defined in socket.c instead of .h, so we hardcode its
+ * value here.
+ */
+ ASSERT_ON_COMPILE(sizeof *vmci_addr <= 128);
+ memcpy(addr, vmci_addr, sizeof *vmci_addr);
+ *addr_len = sizeof *vmci_addr;
+
+out:
+ release_sock(sk);
+ return err;
+}
+
+/*
+ * vsock_vmci_poll --
+ *
+ * Waits on file for activity then provides mask indicating state of socket.
+ *
+ * Results: Mask of flags containing socket state.
+ *
+ * Side effects: None.
+ */
+
+static unsigned int
+vsock_vmci_poll(struct file *file, struct socket *sock, poll_table *wait)
+{
+ struct sock *sk;
+ unsigned int mask;
+ vsock_vmci_sock *vsk;
+
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+
+ poll_wait(file, sk_sleep(sk), wait);
+ mask = 0;
+
+ if (sk->sk_err)
+ /* Signify that there has been an error on this socket. */
+ mask |= POLLERR;
+
+ /*
+ * INET sockets treat local write shutdown and peer write shutdown as a
+ * case of POLLHUP set.
+ */
+ if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
+ ((sk->sk_shutdown & SEND_SHUTDOWN) &&
+ (vsk->peer_shutdown & SEND_SHUTDOWN))) {
+ mask |= POLLHUP;
+ }
+
+ /* POLLRDHUP wasn't added until 2.6.17. */
+ if (sk->sk_shutdown & RCV_SHUTDOWN ||
+ vsk->peer_shutdown & SEND_SHUTDOWN) {
+ mask |= POLLRDHUP;
+ }
+
+ if (sock->type == SOCK_DGRAM) {
+ /*
+ * For datagram sockets we can read if there is something in
+ * the queue and write as long as the socket isn't shutdown for
+ * sending.
+ */
+ if (!skb_queue_empty(&sk->sk_receive_queue) ||
+ (sk->sk_shutdown & RCV_SHUTDOWN)) {
+ mask |= POLLIN | POLLRDNORM;
+ }
+
+ if (!(sk->sk_shutdown & SEND_SHUTDOWN))
+ mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
+
+ } else if (sock->type == SOCK_STREAM) {
+ lock_sock(sk);
+
+ /*
+ * Listening sockets that have connections in their accept
+ * queue can be read.
+ */
+ if (sk->sk_state == SS_LISTEN
+ && !vsock_vmci_is_accept_queue_empty(sk))
+ mask |= POLLIN | POLLRDNORM;
+
+ /*
+ * If there is something in the queue then we can read.
+ */
+ if (!VMCI_HANDLE_INVALID(vsk->qp_handle) &&
+ !(sk->sk_shutdown & RCV_SHUTDOWN)) {
+ bool data_ready_now = false;
+ int ret = 0;
+ NOTIFYCALLRET(vsk, ret, poll_in, sk, 1,
+ &data_ready_now);
+ if (ret < 0) {
+ mask |= POLLERR;
+ } else {
+ if (data_ready_now)
+ mask |= POLLIN | POLLRDNORM;
+
+ }
+ }
+
+ /*
+ * Sockets whose connections have been closed, reset, or
+ * terminated should also be considered read, and we check the
+ * shutdown flag for that.
+ */
+ if (sk->sk_shutdown & RCV_SHUTDOWN ||
+ vsk->peer_shutdown & SEND_SHUTDOWN) {
+ mask |= POLLIN | POLLRDNORM;
+ }
+
+ /*
+ * Connected sockets that can produce data can be written.
+ */
+ if (sk->sk_state == SS_CONNECTED) {
+ if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
+ bool space_avail_now = false;
+ int ret = 0;
+
+ NOTIFYCALLRET(vsk, ret, poll_out, sk, 1,
+ &space_avail_now);
+ if (ret < 0) {
+ mask |= POLLERR;
+ } else {
+ if (space_avail_now)
+ /*
+ * Remove POLLWRBAND since INET
+ * sockets are not setting it.
+ */
+ mask |= POLLOUT | POLLWRNORM;
+
+ }
+ }
+ }
+
+ /*
+ * Simulate INET socket poll behaviors, which sets
+ * POLLOUT|POLLWRNORM when peer is closed and nothing to read,
+ * but local send is not shutdown.
+ */
+ if (sk->sk_state == SS_UNCONNECTED) {
+ if (!(sk->sk_shutdown & SEND_SHUTDOWN))
+ mask |= POLLOUT | POLLWRNORM;
+
+ }
+
+ release_sock(sk);
+ }
+
+ return mask;
+}
+
+/*
+ * vsock_vmci_listen --
+ *
+ * Signify that this socket is listening for connection requests.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_listen(struct socket *sock, int backlog)
+{
+ int err;
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+
+ sk = sock->sk;
+
+ lock_sock(sk);
+
+ if (sock->type != SOCK_STREAM) {
+ err = -EOPNOTSUPP;
+ goto out;
+ }
+
+ if (sock->state != SS_UNCONNECTED) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ vsk = vsock_sk(sk);
+
+ if (!vsock_addr_bound(&vsk->local_addr)) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ sk->sk_max_ack_backlog = backlog;
+ sk->sk_state = SS_LISTEN;
+
+ err = 0;
+
+out:
+ release_sock(sk);
+ return err;
+}
+
+/*
+ * vsock_vmci_shutdown --
+ *
+ * Shuts down the provided socket in the provided method.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_shutdown(struct socket *sock, int mode)
+{
+ int err;
+ struct sock *sk;
+
+ /*
+ * User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
+ * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
+ * here like the other address families do. Note also that the
+ * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
+ * which is what we want.
+ */
+ mode++;
+
+ if ((mode & ~SHUTDOWN_MASK) || !mode)
+ return -EINVAL;
+
+ /*
+ * If this is a STREAM socket and it is not connected then bail out
+ * immediately. If it is a DGRAM socket then we must first kick the socket
+ * so that it wakes up from any sleeping calls, for example recv(), and then
+ * afterwards return the error.
+ */
+
+ sk = sock->sk;
+ if (sock->state == SS_UNCONNECTED) {
+ err = -ENOTCONN;
+ if (sk->sk_type == SOCK_STREAM)
+ return err;
+ } else {
+ sock->state = SS_DISCONNECTING;
+ err = 0;
+ }
+
+ /* Receive and send shutdowns are treated alike. */
+ mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
+ if (mode) {
+ lock_sock(sk);
+ sk->sk_shutdown |= mode;
+ sk->sk_state_change(sk);
+ release_sock(sk);
+
+ if (sk->sk_type == SOCK_STREAM) {
+ sock_reset_flag(sk, SOCK_DONE);
+ VSOCK_SEND_SHUTDOWN(sk, mode);
+ }
+ }
+
+ return err;
+}
+
+/*
+ * vsock_vmci_dgram_sendmsg --
+ *
+ * Sends a datagram.
+ *
+ * Results: Number of bytes sent on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_dgram_sendmsg(struct kiocb *kiocb,
+ struct socket *sock, struct msghdr *msg, size_t len)
+{
+ int err;
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+ struct sockaddr_vm *remote_addr;
+ struct vmci_datagram *dg;
+
+ if (msg->msg_flags & MSG_OOB)
+ return -EOPNOTSUPP;
+
+ if (len > VMCI_MAX_DG_PAYLOAD_SIZE)
+ return -EMSGSIZE;
+
+ /* For now, MSG_DONTWAIT is always assumed... */
+ err = 0;
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+
+ lock_sock(sk);
+
+ if (!vsock_addr_bound(&vsk->local_addr)) {
+ struct sockaddr_vm local_addr;
+
+ vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+ err = __vsock_vmci_bind(sk, &local_addr);
+ if (err != 0)
+ goto out;
+
+ }
+
+ /*
+ * If the provided message contains an address, use that. Otherwise
+ * fall back on the socket's remote handle (if it has been connected).
+ */
+ if (msg->msg_name &&
+ vsock_addr_cast(msg->msg_name, msg->msg_namelen,
+ &remote_addr) == 0) {
+ /* Ensure this address is of the right type and is a valid
+ * destination. */
+
+ if (remote_addr->svm_cid == VMADDR_CID_ANY)
+ remote_addr->svm_cid = vmci_get_context_id();
+
+ if (!vsock_addr_bound(remote_addr)) {
+ err = -EINVAL;
+ goto out;
+ }
+ } else if (sock->state == SS_CONNECTED) {
+ remote_addr = &vsk->remote_addr;
+
+ if (remote_addr->svm_cid == VMADDR_CID_ANY)
+ remote_addr->svm_cid = vmci_get_context_id();
+
+ /* XXX Should connect() or this function ensure remote_addr is
+ * bound? */
+ if (!vsock_addr_bound(&vsk->remote_addr)) {
+ err = -EINVAL;
+ goto out;
+ }
+ } else {
+ err = -EINVAL;
+ goto out;
+ }
+
+ /*
+ * Make sure that we don't allow a userlevel app to send datagrams to
+ * the hypervisor that modify VMCI device state.
+ */
+ if (!vsock_addr_socket_context_dgram(remote_addr->svm_cid,
+ remote_addr->svm_port)) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ if (!vsock_vmci_allow_dgram(vsk, remote_addr->svm_cid)) {
+ err = -EPERM;
+ goto out;
+ }
+
+ /*
+ * Allocate a buffer for the user's message and our packet header.
+ */
+ dg = kmalloc(len + sizeof *dg, GFP_KERNEL);
+ if (!dg) {
+ err = -ENOMEM;
+ goto out;
+ }
+
+ memcpy_fromiovec(VMCI_DG_PAYLOAD(dg), msg->msg_iov, len);
+
+ dg->dst = VMCI_MAKE_HANDLE(remote_addr->svm_cid, remote_addr->svm_port);
+ dg->src =
+ VMCI_MAKE_HANDLE(vsk->local_addr.svm_cid, vsk->local_addr.svm_port);
+
+ dg->payload_size = len;
+
+ err = vmci_datagram_send(dg);
+ kfree(dg);
+ if (err < 0) {
+ err = vsock_vmci_error_to_vsock_error(err);
+ goto out;
+ }
+
+ err -= sizeof *dg;
+
+out:
+ release_sock(sk);
+ return err;
+}
+
+/*
+ * vsock_vmci_stream_setsockopt --
+ *
+ * Set a socket option on a stream socket
+ *
+ * Results: 0 on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_stream_setsockopt(struct socket *sock,
+ int level,
+ int optname,
+ char __user *optval,
+ vsock_setsockopt_len_type optlen)
+{
+ int err;
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+ u64 val;
+
+ if (level != AF_VSOCK)
+ return -ENOPROTOOPT;
+
+#define COPY_IN(_v) \
+ do { \
+ if (optlen < sizeof _v) { \
+ err = -EINVAL; \
+ goto exit; \
+ } \
+ if (copy_from_user(&_v, optval, sizeof _v) != 0) { \
+ err = -EFAULT; \
+ goto exit; \
+ } \
+ } while (0)
+
+ err = 0;
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+
+ ASSERT(vsk->queue_pair_min_size <= vsk->queue_pair_size &&
+ vsk->queue_pair_size <= vsk->queue_pair_max_size);
+
+ lock_sock(sk);
+
+ switch (optname) {
+ case SO_VMCI_BUFFER_SIZE:
+ COPY_IN(val);
+ if (val < vsk->queue_pair_min_size)
+ vsk->queue_pair_min_size = val;
+
+ if (val > vsk->queue_pair_max_size)
+ vsk->queue_pair_max_size = val;
+
+ vsk->queue_pair_size = val;
+ break;
+
+ case SO_VMCI_BUFFER_MAX_SIZE:
+ COPY_IN(val);
+ if (val < vsk->queue_pair_size)
+ vsk->queue_pair_size = val;
+
+ vsk->queue_pair_max_size = val;
+ break;
+
+ case SO_VMCI_BUFFER_MIN_SIZE:
+ COPY_IN(val);
+ if (val > vsk->queue_pair_size)
+ vsk->queue_pair_size = val;
+
+ vsk->queue_pair_min_size = val;
+ break;
+
+ case SO_VMCI_CONNECT_TIMEOUT: {
+ struct timeval tv;
+ COPY_IN(tv);
+ if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
+ tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
+ vsk->connect_timeout = tv.tv_sec * HZ +
+ DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
+ if (vsk->connect_timeout == 0)
+ vsk->connect_timeout =
+ VSOCK_DEFAULT_CONNECT_TIMEOUT;
+
+ } else {
+ err = -ERANGE;
+ }
+ break;
+ }
+
+ default:
+ err = -ENOPROTOOPT;
+ break;
+ }
+
+#undef COPY_IN
+
+ ASSERT(vsk->queue_pair_min_size <= vsk->queue_pair_size &&
+ vsk->queue_pair_size <= vsk->queue_pair_max_size);
+exit:
+ release_sock(sk);
+ return err;
+}
+
+/*
+ * vsock_vmci_stream_getsockopt --
+ *
+ * Get a socket option for a stream socket
+ *
+ * Results: 0 on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_stream_getsockopt(struct socket *sock,
+ int level, int optname,
+ char __user *optval,
+ int __user *optlen)
+{
+ int err;
+ int len;
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+
+ if (level != AF_VSOCK)
+ return -ENOPROTOOPT;
+
+ err = get_user(len, optlen);
+ if (err != 0)
+ return err;
+
+#define COPY_OUT(_v) \
+ do { \
+ if (len < sizeof _v) \
+ return -EINVAL; \
+ \
+ len = sizeof _v; \
+ if (copy_to_user(optval, &_v, len) != 0) \
+ return -EFAULT; \
+ \
+ } while (0)
+
+ err = 0;
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+
+ switch (optname) {
+ case SO_VMCI_BUFFER_SIZE:
+ COPY_OUT(vsk->queue_pair_size);
+ break;
+
+ case SO_VMCI_BUFFER_MAX_SIZE:
+ COPY_OUT(vsk->queue_pair_max_size);
+ break;
+
+ case SO_VMCI_BUFFER_MIN_SIZE:
+ COPY_OUT(vsk->queue_pair_min_size);
+ break;
+
+ case SO_VMCI_CONNECT_TIMEOUT: {
+ struct timeval tv;
+ tv.tv_sec = vsk->connect_timeout / HZ;
+ tv.tv_usec =
+ (vsk->connect_timeout -
+ tv.tv_sec * HZ) * (1000000 / HZ);
+ COPY_OUT(tv);
+ break;
+ }
+ default:
+ return -ENOPROTOOPT;
+ }
+
+ err = put_user(len, optlen);
+ if (err != 0)
+ return -EFAULT;
+
+#undef COPY_OUT
+
+ return 0;
+}
+
+/*
+ * vsock_vmci_stream_sendmsg --
+ *
+ * Sends a message on the socket.
+ *
+ * Results: Number of bytes sent on success, negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_stream_sendmsg(struct kiocb *kiocb,
+ struct socket *sock, struct msghdr *msg, size_t len)
+{
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+ ssize_t total_written;
+ long timeout;
+ int err;
+ vsock_vmci_send_notify_data send_data;
+
+ DEFINE_WAIT(wait);
+
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+ total_written = 0;
+ err = 0;
+
+ if (msg->msg_flags & MSG_OOB)
+ return -EOPNOTSUPP;
+
+ lock_sock(sk);
+
+ /* Callers should not provide a destination with stream sockets. */
+ if (msg->msg_namelen) {
+ err = sk->sk_state == SS_CONNECTED ? -EISCONN : -EOPNOTSUPP;
+ goto out;
+ }
+
+ /* Send data only if both sides are not shutdown in the direction. */
+ if (sk->sk_shutdown & SEND_SHUTDOWN ||
+ vsk->peer_shutdown & RCV_SHUTDOWN) {
+ err = -EPIPE;
+ goto out;
+ }
+
+ if (sk->sk_state != SS_CONNECTED ||
+ !vsock_addr_bound(&vsk->local_addr)) {
+ err = -ENOTCONN;
+ goto out;
+ }
+
+ if (!vsock_addr_bound(&vsk->remote_addr)) {
+ err = -EDESTADDRREQ;
+ goto out;
+ }
+
+ /*
+ * Wait for room in the produce queue to enqueue our user's data.
+ */
+ timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
+
+ NOTIFYCALLRET(vsk, err, send_init, sk, &send_data);
+ if (err < 0)
+ goto out;
+
+ prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
+
+ while (total_written < len) {
+ ssize_t written;
+
+ while (vsock_vmci_stream_has_space(vsk) == 0 &&
+ sk->sk_err == 0 &&
+ !(sk->sk_shutdown & SEND_SHUTDOWN) &&
+ !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
+
+ /* Don't wait for non-blocking sockets. */
+ if (timeout == 0) {
+ err = -EAGAIN;
+ goto out_wait;
+ }
+
+ NOTIFYCALLRET(vsk, err, send_pre_block, sk, &send_data);
+
+ if (err < 0)
+ goto out_wait;
+
+ release_sock(sk);
+ timeout = schedule_timeout(timeout);
+ lock_sock(sk);
+ if (signal_pending(current)) {
+ err = sock_intr_errno(timeout);
+ goto out_wait;
+ } else if (timeout == 0) {
+ err = -EAGAIN;
+ goto out_wait;
+ }
+
+ prepare_to_wait(sk_sleep(sk), &wait,
+ TASK_INTERRUPTIBLE);
+ }
+
+ /*
+ * These checks occur both as part of and after the loop
+ * conditional since we need to check before and after
+ * sleeping.
+ */
+ if (sk->sk_err) {
+ err = -sk->sk_err;
+ goto out_wait;
+ } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
+ (vsk->peer_shutdown & RCV_SHUTDOWN)) {
+ err = -EPIPE;
+ goto out_wait;
+ }
+
+ VSOCK_STATS_STREAM_PRODUCE_HIST(vsk);
+
+ NOTIFYCALLRET(vsk, err, send_pre_enqueue, sk, &send_data);
+ if (err < 0)
+ goto out_wait;
+
+ /*
+ * Note that enqueue will only write as many bytes as are free
+ * in the produce queue, so we don't need to ensure len is
+ * smaller than the queue size. It is the caller's
+ * responsibility to check how many bytes we were able to send.
+ */
+
+ written = vmci_qpair_enquev(vsk->qpair, msg->msg_iov,
+ len - total_written, 0);
+ if (written < 0) {
+ err = -ENOMEM;
+ goto out_wait;
+ }
+
+ total_written += written;
+
+ NOTIFYCALLRET(vsk, err, send_post_enqueue, sk, written,
+ &send_data);
+ if (err < 0)
+ goto out_wait;
+
+ }
+
+ ASSERT(total_written <= INT_MAX);
+
+out_wait:
+ if (total_written > 0) {
+ VSOCK_STATS_STREAM_PRODUCE(total_written);
+ err = total_written;
+ }
+ finish_wait(sk_sleep(sk), &wait);
+out:
+ release_sock(sk);
+ return err;
+}
+
+/*
+ * vsock_vmci_dgram_recvmsg --
+ *
+ * Receives a datagram and places it in the caller's msg.
+ *
+ * Results: The size of the payload on success, negative value on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_dgram_recvmsg(struct kiocb *kiocb,
+ struct socket *sock,
+ struct msghdr *msg, size_t len, int flags)
+{
+ int err;
+ int noblock;
+ struct sock *sk;
+ struct vmci_datagram *dg;
+ size_t payload_len;
+ struct sk_buff *skb;
+
+ sk = sock->sk;
+ noblock = flags & MSG_DONTWAIT;
+
+ if (flags & MSG_OOB || flags & MSG_ERRQUEUE)
+ return -EOPNOTSUPP;
+
+ /* Retrieve the head sk_buff from the socket's receive queue. */
+ err = 0;
+ skb = skb_recv_datagram(sk, flags, noblock, &err);
+ if (err)
+ return err;
+
+ if (!skb)
+ return -EAGAIN;
+
+ dg = (struct vmci_datagram *) skb->data;
+ if (!dg)
+ /* err is 0, meaning we read zero bytes. */
+ goto out;
+
+ payload_len = dg->payload_size;
+ /* Ensure the sk_buff matches the payload size claimed in the packet. */
+ if (payload_len != skb->len - sizeof *dg) {
+ err = -EINVAL;
+ goto out;
+ }
+
+ if (payload_len > len) {
+ payload_len = len;
+ msg->msg_flags |= MSG_TRUNC;
+ }
+
+ /* Place the datagram payload in the user's iovec. */
+ err =
+ skb_copy_datagram_iovec(skb, sizeof *dg, msg->msg_iov, payload_len);
+ if (err)
+ goto out;
+
+ msg->msg_namelen = 0;
+ if (msg->msg_name) {
+ struct sockaddr_vm *vmci_addr;
+
+ /* Provide the address of the sender. */
+ vmci_addr = (struct sockaddr_vm *)msg->msg_name;
+ vsock_addr_init(vmci_addr,
+ VMCI_HANDLE_TO_CONTEXT_ID(dg->src),
+ VMCI_HANDLE_TO_RESOURCE_ID(dg->src));
+ msg->msg_namelen = sizeof *vmci_addr;
+ }
+ err = payload_len;
+
+out:
+ skb_free_datagram(sk, skb);
+ return err;
+}
+
+/*
+ * vsock_vmci_stream_recvmsg --
+ *
+ * Receives a datagram and places it in the caller's msg.
+ *
+ * Results: The size of the payload on success, negative value on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_stream_recvmsg(struct kiocb *kiocb,
+ struct socket *sock,
+ struct msghdr *msg, size_t len, int flags)
+{
+ struct sock *sk;
+ vsock_vmci_sock *vsk;
+ int err;
+ size_t target;
+ ssize_t copied;
+ long timeout;
+
+ vsock_vmci_recv_notify_data recv_data;
+
+ DEFINE_WAIT(wait);
+
+ sk = sock->sk;
+ vsk = vsock_sk(sk);
+ err = 0;
+
+ lock_sock(sk);
+
+ if (sk->sk_state != SS_CONNECTED) {
+ /*
+ * Recvmsg is supposed to return 0 if a peer performs an
+ * orderly shutdown. Differentiate between that case and when a
+ * peer has not connected or a local shutdown occured with the
+ * SOCK_DONE flag.
+ */
+ if (sock_flag(sk, SOCK_DONE))
+ err = 0;
+ else
+ err = -ENOTCONN;
+
+ goto out;
+ }
+
+ if (flags & MSG_OOB) {
+ err = -EOPNOTSUPP;
+ goto out;
+ }
+
+ /*
+ * We don't check peer_shutdown flag here since peer may actually shut
+ * down, but there can be data in the VMCI queue that local socket can
+ * receive.
+ */
+ if (sk->sk_shutdown & RCV_SHUTDOWN) {
+ err = 0;
+ goto out;
+ }
+
+ /*
+ * It is valid on Linux to pass in a zero-length receive buffer. This
+ * is not an error. We may as well bail out now. Note that if we
+ * don't, we will fail "ASSERT(copied >= target)" after we dequeue,
+ * because the minimum target is always 1 byte.
+ */
+ if (!len) {
+ err = 0;
+ goto out;
+ }
+
+ /*
+ * We must not copy less than target bytes into the user's buffer
+ * before returning successfully, so we wait for the consume queue to
+ * have that much data to consume before dequeueing. Note that this
+ * makes it impossible to handle cases where target is greater than the
+ * queue size.
+ */
+ target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
+ if (target >= vsk->consume_size) {
+ err = -ENOMEM;
+ goto out;
+ }
+ timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
+ copied = 0;
+
+ NOTIFYCALLRET(vsk, err, recv_init, sk, target, &recv_data);
+ if (err < 0)
+ goto out;
+
+ prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
+
+ while (1) {
+ s64 ready = vsock_vmci_stream_has_data(vsk);
+
+ if (ready < 0) {
+ /*
+ * Invalid queue pair content. XXX This should be
+ * changed to a connection reset in a later change.
+ */
+
+ err = -ENOMEM;
+ goto out_wait;
+ } else if (ready > 0) {
+ ssize_t read;
+
+ VSOCK_STATS_STREAM_CONSUME_HIST(vsk);
+
+ NOTIFYCALLRET(vsk, err, recv_pre_dequeue, sk, target,
+ &recv_data);
+ if (err < 0)
+ break;
+
+ if (flags & MSG_PEEK)
+ read =
+ vmci_qpair_peekv(vsk->qpair, msg->msg_iov,
+ len - copied, 0);
+ else
+ read =
+ vmci_qpair_dequev(vsk->qpair, msg->msg_iov,
+ len - copied, 0);
+
+ if (read < 0) {
+ err = -ENOMEM;
+ break;
+ }
+
+ ASSERT(read <= INT_MAX);
+ copied += read;
+
+ NOTIFYCALLRET(vsk, err, recv_post_dequeue, sk, target,
+ read, !(flags & MSG_PEEK), &recv_data);
+ if (err < 0)
+ goto out_wait;
+
+ if (read >= target || flags & MSG_PEEK)
+ break;
+
+ target -= read;
+ } else {
+ if (sk->sk_err != 0 || (sk->sk_shutdown & RCV_SHUTDOWN)
+ || (vsk->peer_shutdown & SEND_SHUTDOWN)) {
+ break;
+ }
+ /* Don't wait for non-blocking sockets. */
+ if (timeout == 0) {
+ err = -EAGAIN;
+ break;
+ }
+
+ NOTIFYCALLRET(vsk, err, recv_pre_block, sk, target,
+ &recv_data);
+ if (err < 0)
+ break;
+
+ release_sock(sk);
+ timeout = schedule_timeout(timeout);
+ lock_sock(sk);
+
+ if (signal_pending(current)) {
+ err = sock_intr_errno(timeout);
+ break;
+ } else if (timeout == 0) {
+ err = -EAGAIN;
+ break;
+ }
+
+ prepare_to_wait(sk_sleep(sk), &wait,
+ TASK_INTERRUPTIBLE);
+ }
+ }
+
+ if (sk->sk_err)
+ err = -sk->sk_err;
+ else if (sk->sk_shutdown & RCV_SHUTDOWN)
+ err = 0;
+
+ if (copied > 0) {
+ /*
+ * We only do these additional bookkeeping/notification steps
+ * if we actually copied something out of the queue pair
+ * instead of just peeking ahead.
+ */
+
+ if (!(flags & MSG_PEEK)) {
+ VSOCK_STATS_STREAM_CONSUME(copied);
+
+ /*
+ * If the other side has shutdown for sending and there
+ * is nothing more to read, then modify the socket
+ * state.
+ */
+ if (vsk->peer_shutdown & SEND_SHUTDOWN) {
+ if (vsock_vmci_stream_has_data(vsk) <= 0) {
+ sk->sk_state = SS_UNCONNECTED;
+ sock_set_flag(sk, SOCK_DONE);
+ sk->sk_state_change(sk);
+ }
+ }
+ }
+ err = copied;
+ }
+
+out_wait:
+ finish_wait(sk_sleep(sk), &wait);
+out:
+ release_sock(sk);
+ return err;
+}
+
+/*
+ * Protocol operation.
+ */
+
+/*
+ * vsock_vmci_create --
+ *
+ * Creates a VSocket socket.
+ *
+ * Results: Zero on success, negative error code on failure.
+ *
+ * Side effects: Socket count is incremented.
+ */
+
+static int
+vsock_vmci_create(struct net *net, struct socket *sock, int protocol, int kern)
+{
+ if (!sock)
+ return -EINVAL;
+
+ if (protocol)
+ return -EPROTONOSUPPORT;
+
+ switch (sock->type) {
+ case SOCK_DGRAM:
+ sock->ops = &vsock_vmci_dgram_ops;
+ break;
+ case SOCK_STREAM:
+ sock->ops = &vsock_vmci_stream_ops;
+ break;
+ default:
+ return -ESOCKTNOSUPPORT;
+ }
+
+ sock->state = SS_UNCONNECTED;
+
+ return __vsock_vmci_create(net, sock, NULL, GFP_KERNEL,
+ 0) ? 0 : -ENOMEM;
+}
+
+/*
+ * Device operations.
+ */
+
+static long vsock_vmci_dev_do_ioctl(struct file *filp,
+ unsigned int cmd, void __user *ptr)
+{
+ static const u16 parts[4] = { VSOCK_DRIVER_VERSION_COMMAS };
+ u32 __user *p = ptr;
+ int retval = 0;
+ u32 version;
+
+ switch (cmd) {
+ case IOCTL_VMCI_SOCKETS_VERSION:
+ version = VMCI_SOCKETS_MAKE_VERSION(parts);
+ if (put_user(version, p) != 0)
+ retval = -EFAULT;
+ break;
+
+ case IOCTL_VMCI_SOCKETS_GET_AF_VALUE:
+ if (put_user(AF_VSOCK, p) != 0)
+ retval = -EFAULT;
+
+ break;
+
+ case IOCTL_VMCI_SOCKETS_GET_LOCAL_CID:
+ if (put_user(vmci_get_context_id(), p) != 0)
+ retval = -EFAULT;
+
+ break;
+
+ default:
+ pr_err("Unknown ioctl %d\n", cmd);
+ retval = -EINVAL;
+ }
+
+ return retval;
+}
+
+static long vsock_vmci_dev_ioctl(struct file *filp,
+ unsigned int cmd, unsigned long arg)
+{
+ return vsock_vmci_dev_do_ioctl(filp, cmd, (void __user *)arg);
+}
+
+#ifdef CONFIG_COMPAT
+static long vsock_vmci_dev_compat_ioctl(struct file *filp,
+ unsigned int cmd, unsigned long arg)
+{
+ return vsock_vmci_dev_do_ioctl(filp, cmd, compat_ptr(arg));
+}
+#endif
+
+static const struct file_operations vsock_vmci_device_ops = {
+ .owner = THIS_MODULE,
+ .unlocked_ioctl = vsock_vmci_dev_ioctl,
+#ifdef CONFIG_COMPAT
+ .compat_ioctl = vsock_vmci_dev_compat_ioctl,
+#endif
+ .open = nonseekable_open,
+};
+
+static struct miscdevice vsock_vmci_device = {
+ .name = "vsock",
+ .minor = MISC_DYNAMIC_MINOR,
+ .fops = &vsock_vmci_device_ops,
+};
+
+
+/*
+ * Module operations.
+ */
+
+/*
+ * vsock_vmci_init --
+ *
+ * Initialization routine for the VSockets module.
+ *
+ * Results: Zero on success, error code on failure.
+ *
+ * Side effects: The VSocket protocol family and socket operations are
+ * registered.
+ */
+
+static int __init vsock_vmci_init(void)
+{
+ int err;
+
+ request_module("vmci");
+
+ err = misc_register(&vsock_vmci_device);
+ if (err) {
+ pr_err("Failed to register misc device\n");
+ return -ENOENT;
+ }
+
+ err = vsock_vmci_register_with_vmci();
+ if (err) {
+ pr_err("Cannot register with VMCI device.\n");
+ goto err_misc_deregister;
+ }
+
+ err = proto_register(&vsock_vmci_proto, 1); /* we want our slab */
+ if (err) {
+ pr_err("Cannot register vsock protocol.\n");
+ goto err_unregister_with_vmci;
+ }
+
+ err = sock_register(&vsock_vmci_family_ops);
+ if (err) {
+ pr_err("could not register af_vsock (%d) address family: %d\n",
+ AF_VSOCK, err);
+ goto err_unregister_proto;
+ }
+
+ vsock_vmci_init_tables();
+ return 0;
+
+err_unregister_proto:
+ proto_unregister(&vsock_vmci_proto);
+err_unregister_with_vmci:
+ vsock_vmci_unregister_with_vmci();
+err_misc_deregister:
+ misc_deregister(&vsock_vmci_device);
+ return err;
+}
+
+/*
+ * VSocketVmciExit --
+ *
+ * VSockets module exit routine.
+ *
+ * Results: None.
+ *
+ * Side effects: Unregisters VSocket protocol family and socket operations.
+ */
+
+static void __exit vsock_vmci_exit(void)
+{
+ misc_deregister(&vsock_vmci_device);
+ sock_unregister(AF_VSOCK);
+ proto_unregister(&vsock_vmci_proto);
+ /* Need reset ? */
+ VSOCK_STATS_RESET();
+ vsock_vmci_unregister_with_vmci();
+}
+
+module_init(vsock_vmci_init);
+module_exit(vsock_vmci_exit);
+
+MODULE_AUTHOR("VMware, Inc.");
+MODULE_DESCRIPTION("VMware Virtual Socket Family");
+MODULE_VERSION(VSOCK_DRIVER_VERSION_STRING);
+MODULE_LICENSE("GPL v2");
+MODULE_ALIAS("vmware_vsock");
diff --git a/net/vmw_vsock/af_vsock.h b/net/vmw_vsock/af_vsock.h
new file mode 100644
index 0000000..c434afc
--- /dev/null
+++ b/net/vmw_vsock/af_vsock.h
@@ -0,0 +1,179 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * af_vsock.h --
+ *
+ * Definitions for Linux VSockets module.
+ */
+
+#ifndef __AF_VSOCK_H__
+#define __AF_VSOCK_H__
+
+#include <linux/kernel.h>
+#include <linux/workqueue.h>
+#include <linux/vmw_vmci_defs.h>
+#include <linux/vmw_vmci_api.h>
+
+#include "vsock_common.h"
+#include "vsock_packet.h"
+#include "notify.h"
+
+#define vsock_sk(__sk) ((vsock_vmci_sock *)__sk)
+#define sk_vsock(__vsk) (&(__vsk)->sk)
+
+typedef struct vsock_vmci_sock {
+ /* sk must be the first member. */
+ struct sock sk;
+ struct sockaddr_vm local_addr;
+ struct sockaddr_vm remote_addr;
+ /* Links for the global tables of bound and connected sockets. */
+ struct list_head bound_table;
+ struct list_head connected_table;
+ /*
+ * Accessed without the socket lock held. This means it can never be
+ * modified outsided of socket create or destruct.
+ */
+ bool trusted;
+ bool cached_peer_allow_dgram; /* Dgram communication allowed to
+ * cached peer? */
+ vmci_id cached_peer; /* Context ID of last dgram destination check. */
+ const struct cred *owner;
+ struct vmci_handle dg_handle; /* For SOCK_DGRAM only. */
+ /* Rest are SOCK_STREAM only. */
+ struct vmci_handle qp_handle;
+ struct vmci_qp *qpair;
+ u64 produce_size;
+ u64 consume_size;
+ u64 queue_pair_size;
+ u64 queue_pair_min_size;
+ u64 queue_pair_max_size;
+ long connect_timeout;
+ vsock_vmci_notify notify;
+ vsock_vmci_notify_ops *notify_ops;
+ vmci_id attach_sub_id;
+ vmci_id detach_sub_id;
+ /* Listening socket that this came from. */
+ struct sock *listener;
+ /*
+ * Used for pending list and accept queue during connection handshake.
+ * The listening socket is the head for both lists. Sockets created
+ * for connection requests are placed in the pending list until they
+ * are connected, at which point they are put in the accept queue list
+ * so they can be accepted in accept(). If accept() cannot accept the
+ * connection, it is marked as rejected so the cleanup function knows
+ * to clean up the socket.
+ */
+ struct list_head pending_links;
+ struct list_head accept_queue;
+ bool rejected;
+ struct delayed_work dwork;
+ u32 peer_shutdown;
+ bool sent_request;
+ bool ignore_connecting_rst;
+} vsock_vmci_sock;
+
+int vsock_vmci_send_control_pkt_bh(struct sockaddr_vm *src,
+ struct sockaddr_vm *dst,
+ vsock_packet_type type,
+ u64 size,
+ u64 mode,
+ vsock_waiting_info *wait,
+ struct vmci_handle handle);
+int vsock_vmci_reply_control_pkt_fast(vsock_packet *pkt,
+ vsock_packet_type type, u64 size,
+ u64 mode, vsock_waiting_info *wait,
+ struct vmci_handle handle);
+int vsock_vmci_send_control_pkt(struct sock *sk, vsock_packet_type type,
+ u64 size, u64 mode,
+ vsock_waiting_info *wait,
+ vsock_proto_version version,
+ struct vmci_handle handle);
+
+s64 vsock_vmci_stream_has_data(vsock_vmci_sock *vsk);
+s64 vsock_vmci_stream_has_space(vsock_vmci_sock *vsk);
+
+#define VSOCK_SEND_RESET_BH(_dst, _src, _pkt) \
+ ((_pkt)->type == VSOCK_PACKET_TYPE_RST) ? \
+ 0 : \
+ vsock_vmci_send_control_pkt_bh( \
+ _dst, _src, \
+ VSOCK_PACKET_TYPE_RST, 0, \
+ 0, NULL, VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_INVALID_BH(_dst, _src) \
+ vsock_vmci_send_control_pkt_bh(_dst, _src, \
+ VSOCK_PACKET_TYPE_INVALID, 0, \
+ 0, NULL, VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_WROTE_BH(_dst, _src) \
+ vsock_vmci_send_control_pkt_bh(_dst, _src, VSOCK_PACKET_TYPE_WROTE, 0, \
+ 0, NULL, VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_READ_BH(_dst, _src) \
+ vsock_vmci_send_control_pkt_bh((_dst), (_src), \
+ VSOCK_PACKET_TYPE_READ, 0, \
+ 0, NULL, VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_RESET(_sk, _pkt) \
+ ((_pkt)->type == VSOCK_PACKET_TYPE_RST) ? \
+ 0 : \
+ vsock_vmci_send_control_pkt( \
+ _sk, VSOCK_PACKET_TYPE_RST, \
+ 0, 0, NULL, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_NEGOTIATE(_sk, _size) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_NEGOTIATE, \
+ _size, 0, NULL, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_NEGOTIATE2(_sk, _size, signal_proto) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_NEGOTIATE2, \
+ _size, 0, NULL, signal_proto, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_QP_OFFER(_sk, _handle) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_OFFER, \
+ 0, 0, NULL, VSOCK_PROTO_INVALID, _handle)
+#define VSOCK_SEND_CONN_REQUEST(_sk, _size) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_REQUEST, \
+ _size, 0, NULL, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_CONN_REQUEST2(_sk, _size, signal_proto) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_REQUEST2, \
+ _size, 0, NULL, signal_proto, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_ATTACH(_sk, _handle) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_ATTACH, \
+ 0, 0, NULL, VSOCK_PROTO_INVALID, _handle)
+#define VSOCK_SEND_WROTE(_sk) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_WROTE, \
+ 0, 0, NULL, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_READ(_sk) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_READ, \
+ 0, 0, NULL, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_SHUTDOWN(_sk, _mode) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_SHUTDOWN, \
+ 0, _mode, NULL, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_WAITING_WRITE(_sk, _wait_info) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_WAITING_WRITE, \
+ 0, 0, _wait_info, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_SEND_WAITING_READ(_sk, _wait_info) \
+ vsock_vmci_send_control_pkt(_sk, VSOCK_PACKET_TYPE_WAITING_READ, \
+ 0, 0, _wait_info, VSOCK_PROTO_INVALID, \
+ VMCI_INVALID_HANDLE)
+#define VSOCK_REPLY_RESET(_pkt) \
+ vsock_vmci_reply_control_pkt_fast(_pkt, VSOCK_PACKET_TYPE_RST, \
+ 0, 0, NULL, VMCI_INVALID_HANDLE)
+
+#endif /* __AF_VSOCK_H__ */
VSOCK control notifications for VMCI Stream Sockets protocol.
Signed-off-by: George Zhang <[email protected]>
---
net/vmw_vsock/notify.c | 984 ++++++++++++++++++++++++++++++++++++++++++++++++
net/vmw_vsock/notify.h | 130 ++++++
2 files changed, 1114 insertions(+), 0 deletions(-)
create mode 100644 net/vmw_vsock/notify.c
create mode 100644 net/vmw_vsock/notify.h
diff --git a/net/vmw_vsock/notify.c b/net/vmw_vsock/notify.c
new file mode 100644
index 0000000..5c69c7c
--- /dev/null
+++ b/net/vmw_vsock/notify.c
@@ -0,0 +1,984 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2009-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * notify.c --
+ *
+ * Linux control notifications for the VMCI Stream Sockets protocol.
+ */
+
+#include <linux/types.h>
+
+#include <linux/socket.h>
+#include <linux/stddef.h> /* for NULL */
+#include <net/sock.h>
+
+#include "notify.h"
+#include "af_vsock.h"
+
+#define PKT_FIELD(vsk, field_name) \
+ (vsk)->notify.pkt.field_name
+
+#define VSOCK_MAX_DGRAM_RESENDS 10
+
+/*
+ *
+ * vsock_vmci_notify_waiting_write --
+ *
+ * Determines if the conditions have been met to notify a waiting writer.
+ *
+ * Results: true if a notification should be sent, false otherwise.
+ *
+ * Side effects: None.
+ */
+
+static bool vsock_vmci_notify_waiting_write(vsock_vmci_sock *vsk)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ bool retval;
+ u64 notify_limit;
+
+ if (!PKT_FIELD(vsk, peer_waiting_write))
+ return false;
+
+#ifdef VSOCK_OPTIMIZATION_FLOW_CONTROL
+ /*
+ * When the sender blocks, we take that as a sign that the sender is
+ * faster than the receiver. To reduce the transmit rate of the sender,
+ * we delay the sending of the read notification by decreasing the
+ * write_notify_window. The notification is delayed until the number of
+ * bytes used in the queue drops below the write_notify_window.
+ */
+
+ if (!PKT_FIELD(vsk, peer_waiting_write_detected)) {
+ PKT_FIELD(vsk, peer_waiting_write_detected) = true;
+ if (PKT_FIELD(vsk, write_notify_window) < PAGE_SIZE) {
+ PKT_FIELD(vsk, write_notify_window) =
+ PKT_FIELD(vsk, write_notify_min_window);
+ } else {
+ PKT_FIELD(vsk, write_notify_window) -= PAGE_SIZE;
+ if (PKT_FIELD(vsk, write_notify_window) <
+ PKT_FIELD(vsk, write_notify_min_window))
+ PKT_FIELD(vsk, write_notify_window) =
+ PKT_FIELD(vsk, write_notify_min_window);
+
+ }
+ }
+ notify_limit = vsk->consume_size - PKT_FIELD(vsk, write_notify_window);
+#else
+ notify_limit = 0;
+#endif
+
+ /*
+ * For now we ignore the wait information and just see if the free
+ * space exceeds the notify limit. Note that improving this function
+ * to be more intelligent will not require a protocol change and will
+ * retain compatibility between endpoints with mixed versions of this
+ * function.
+ *
+ * The notify_limit is used to delay notifications in the case where
+ * flow control is enabled. Below the test is expressed in terms of
+ * free space in the queue: if free_space > ConsumeSize -
+ * write_notify_window then notify An alternate way of expressing this
+ * is to rewrite the expression to use the data ready in the receive
+ * queue: if write_notify_window > bufferReady then notify as
+ * free_space == ConsumeSize - bufferReady.
+ */
+ retval = vmci_qpair_consume_free_space(vsk->qpair) > notify_limit;
+#ifdef VSOCK_OPTIMIZATION_FLOW_CONTROL
+ if (retval) {
+ /*
+ * Once we notify the peer, we reset the detected flag so the
+ * next wait will again cause a decrease in the window size.
+ */
+
+ PKT_FIELD(vsk, peer_waiting_write_detected) = false;
+ }
+#endif
+ return retval;
+#else
+ return true;
+#endif
+}
+
+/*
+ *
+ * vsock_vmci_notify_waiting_read --
+ *
+ * Determines if the conditions have been met to notify a waiting reader.
+ *
+ * Results: true if a notification should be sent, false otherwise.
+ *
+ * Side effects: None.
+ */
+
+static bool vsock_vmci_notify_waiting_read(vsock_vmci_sock *vsk)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ if (!PKT_FIELD(vsk, peer_waiting_read))
+ return false;
+
+ /*
+ * For now we ignore the wait information and just see if there is any
+ * data for our peer to read. Note that improving this function to be
+ * more intelligent will not require a protocol change and will retain
+ * compatibility between endpoints with mixed versions of this
+ * function.
+ */
+ return vmci_qpair_produce_buf_ready(vsk->qpair) > 0;
+#else
+ return true;
+#endif
+}
+
+/*
+ *
+ * vsock_vmci_handle_waiting_read --
+ *
+ * Handles an incoming waiting read message.
+ *
+ * Results: None.
+ *
+ * Side effects: May send a notification to the peer, may update socket's wait
+ * info structure.
+ */
+
+static void
+vsock_vmci_handle_waiting_read(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst, struct sockaddr_vm *src)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, peer_waiting_read) = true;
+ memcpy(&PKT_FIELD(vsk, peer_waiting_read_info), &pkt->u.wait,
+ sizeof PKT_FIELD(vsk, peer_waiting_read_info));
+
+ if (vsock_vmci_notify_waiting_read(vsk)) {
+ bool sent;
+
+ if (bottom_half)
+ sent = VSOCK_SEND_WROTE_BH(dst, src) > 0;
+ else
+ sent = VSOCK_SEND_WROTE(sk) > 0;
+
+ if (sent)
+ PKT_FIELD(vsk, peer_waiting_read) = false;
+
+ }
+#endif
+}
+
+/*
+ *
+ * vsock_vmci_handle_waiting_write --
+ *
+ * Handles an incoming waiting write message.
+ *
+ * Results: None.
+ *
+ * Side effects: May send a notification to the peer, may update socket's wait
+ * info structure.
+ */
+
+static void
+vsock_vmci_handle_waiting_write(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst,
+ struct sockaddr_vm *src)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, peer_waiting_write) = true;
+ memcpy(&PKT_FIELD(vsk, peer_waiting_write_info), &pkt->u.wait,
+ sizeof PKT_FIELD(vsk, peer_waiting_write_info));
+
+ if (vsock_vmci_notify_waiting_write(vsk)) {
+ bool sent;
+
+ if (bottom_half)
+ sent = VSOCK_SEND_READ_BH(dst, src) > 0;
+ else
+ sent = VSOCK_SEND_READ(sk) > 0;
+
+ if (sent)
+ PKT_FIELD(vsk, peer_waiting_write) = false;
+
+ }
+#endif
+}
+
+/*
+ *
+ * vsock_vmci_handle_read --
+ *
+ * Handles an incoming read message.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void
+vsock_vmci_handle_read(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst, struct sockaddr_vm *src)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+ PKT_FIELD(vsk, sent_waiting_write) = false;
+#endif
+
+ sk->sk_write_space(sk);
+}
+
+/*
+ *
+ * vsock_vmci_send_waiting_read --
+ *
+ * Sends a waiting read notification to this socket's peer.
+ *
+ * Results: true if the datagram is sent successfully, false otherwise.
+ *
+ * Side effects: Our peer will notify us when there is data to read from our
+ * consume queue.
+ */
+
+static bool vsock_vmci_send_waiting_read(struct sock *sk, u64 room_needed)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vsock_vmci_sock *vsk;
+ vsock_waiting_info waiting_info;
+ u64 tail;
+ u64 head;
+ u64 room_left;
+ bool ret;
+
+ vsk = vsock_sk(sk);
+
+ if (PKT_FIELD(vsk, sent_waiting_read))
+ return true;
+
+ if (PKT_FIELD(vsk, write_notify_window) < vsk->consume_size)
+ PKT_FIELD(vsk, write_notify_window) =
+ min(PKT_FIELD(vsk, write_notify_window) + PAGE_SIZE,
+ vsk->consume_size);
+
+ vmci_qpair_get_consume_indexes(vsk->qpair, &tail, &head);
+ room_left = vsk->consume_size - head;
+ if (room_needed >= room_left) {
+ waiting_info.offset = room_needed - room_left;
+ waiting_info.generation =
+ PKT_FIELD(vsk, consume_q_generation) + 1;
+ } else {
+ waiting_info.offset = head + room_needed;
+ waiting_info.generation = PKT_FIELD(vsk, consume_q_generation);
+ }
+
+ ret = VSOCK_SEND_WAITING_READ(sk, &waiting_info) > 0;
+ if (ret)
+ PKT_FIELD(vsk, sent_waiting_read) = true;
+
+ return ret;
+#else
+ return true;
+#endif
+}
+
+/*
+ *
+ * vsock_vmci_send_waiting_write --
+ *
+ * Sends a waiting write notification to this socket's peer.
+ *
+ * Results: true if the datagram is sent successfully or does not need to be
+ * sent. false otherwise.
+ *
+ * Side effects: Our peer will notify us when there is room to write in to our
+ * produce queue.
+ */
+
+static bool vsock_vmci_send_waiting_write(struct sock *sk, u64 room_needed)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vsock_vmci_sock *vsk;
+ vsock_waiting_info waiting_info;
+ u64 tail;
+ u64 head;
+ u64 room_left;
+ bool ret;
+
+ vsk = vsock_sk(sk);
+
+ if (PKT_FIELD(vsk, sent_waiting_write))
+ return true;
+
+ vmci_qpair_get_produce_indexes(vsk->qpair, &tail, &head);
+ room_left = vsk->produce_size - tail;
+ if (room_needed + 1 >= room_left) {
+ /* Wraps around to current generation. */
+ waiting_info.offset = room_needed + 1 - room_left;
+ waiting_info.generation = PKT_FIELD(vsk, produce_q_generation);
+ } else {
+ waiting_info.offset = tail + room_needed + 1;
+ waiting_info.generation =
+ PKT_FIELD(vsk, produce_q_generation) - 1;
+ }
+
+ ret = VSOCK_SEND_WAITING_WRITE(sk, &waiting_info) > 0;
+ if (ret)
+ PKT_FIELD(vsk, sent_waiting_write) = true;
+
+ return ret;
+#else
+ return true;
+#endif
+}
+
+/*
+ *
+ * vsock_vmci_send_read_notification --
+ *
+ * Sends a read notification to this socket's peer.
+ *
+ * Results: >= 0 if the datagram is sent successfully, negative error value
+ * otherwise.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_send_read_notification(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+ bool sent_read;
+ unsigned int retries;
+ int err;
+
+ vsk = vsock_sk(sk);
+ sent_read = false;
+ retries = 0;
+ err = 0;
+
+ if (vsock_vmci_notify_waiting_write(vsk)) {
+ /*
+ * Notify the peer that we have read, retrying the send on
+ * failure up to our maximum value. XXX For now we just log
+ * the failure, but later we should schedule a work item to
+ * handle the resend until it succeeds. That would require
+ * keeping track of work items in the vsk and cleaning them up
+ * upon socket close.
+ */
+ while (!(vsk->peer_shutdown & RCV_SHUTDOWN) &&
+ !sent_read && retries < VSOCK_MAX_DGRAM_RESENDS) {
+ err = VSOCK_SEND_READ(sk);
+ if (err >= 0)
+ sent_read = true;
+
+ retries++;
+ }
+
+ if (retries >= VSOCK_MAX_DGRAM_RESENDS)
+ printk
+ ("%p unable to send read notify to peer.\n",
+ sk);
+ else
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ PKT_FIELD(vsk, peer_waiting_write) = false;
+#endif
+
+ }
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_handle_wrote --
+ *
+ * Handles an incoming wrote message.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void
+vsock_vmci_handle_wrote(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst, struct sockaddr_vm *src)
+{
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+ PKT_FIELD(vsk, sent_waiting_read) = false;
+#endif
+
+ sk->sk_data_ready(sk, 0);
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_socket_init --
+ *
+ * Function that is called after a socket is created and before any notify ops
+ * are used.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_socket_init(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+ vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, write_notify_window) = PAGE_SIZE;
+ PKT_FIELD(vsk, write_notify_min_window) = PAGE_SIZE;
+ PKT_FIELD(vsk, peer_waiting_read) = false;
+ PKT_FIELD(vsk, peer_waiting_write) = false;
+ PKT_FIELD(vsk, peer_waiting_write_detected) = false;
+ PKT_FIELD(vsk, sent_waiting_read) = false;
+ PKT_FIELD(vsk, sent_waiting_write) = false;
+ PKT_FIELD(vsk, produce_q_generation) = 0;
+ PKT_FIELD(vsk, consume_q_generation) = 0;
+
+ memset(&PKT_FIELD(vsk, peer_waiting_read_info), 0,
+ sizeof PKT_FIELD(vsk, peer_waiting_read_info));
+ memset(&PKT_FIELD(vsk, peer_waiting_write_info), 0,
+ sizeof PKT_FIELD(vsk, peer_waiting_write_info));
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_socket_destruct --
+ *
+ * Function that is called when the socket is being released.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_socket_destruct(struct sock *sk)
+{
+ return;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_poll_in --
+ *
+ * Called by the poll function to figure out if there is data to read and to
+ * setup future notifications if needed. Only called on sockets that aren't
+ * shutdown for recv.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_poll_in(struct sock *sk,
+ size_t target, bool *data_ready_now)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ if (vsock_vmci_stream_has_data(vsk)) {
+ *data_ready_now = true;
+ } else {
+ /*
+ * We can't read right now because there is nothing in the
+ * queue. Ask for notifications when there is something to
+ * read.
+ */
+ if (sk->sk_state == SS_CONNECTED) {
+ if (!vsock_vmci_send_waiting_read(sk, 1))
+ return -1;
+
+ }
+ *data_ready_now = false;
+ }
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_poll_out
+ *
+ * Called by the poll function to figure out if there is space to write and to
+ * setup future notifications if needed. Only called on a connected socket that
+ * isn't shutdown for send.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_poll_out(struct sock *sk,
+ size_t target, bool *space_avail_now)
+{
+ s64 produce_q_free_space;
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ produce_q_free_space = vsock_vmci_stream_has_space(vsk);
+ if (produce_q_free_space > 0) {
+ *space_avail_now = true;
+ return 0;
+ } else if (produce_q_free_space == 0) {
+ /*
+ * This is a connected socket but we can't currently send data.
+ * Notify the peer that we are waiting if the queue is full. We
+ * only send a waiting write if the queue is full because
+ * otherwise we end up in an infinite WAITING_WRITE, READ,
+ * WAITING_WRITE, READ, etc. loop. Treat failing to send the
+ * notification as a socket error, passing that back through
+ * the mask.
+ */
+ if (!vsock_vmci_send_waiting_write(sk, 1))
+ return -1;
+
+ *space_avail_now = false;
+ }
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_recv_init --
+ *
+ * Called at the start of a stream recv call with the socket lock held.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_recv_init(struct sock *sk,
+ size_t target,
+ vsock_vmci_recv_notify_data *data)
+{
+ vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+
+#ifdef VSOCK_OPTIMIZATION_WAITING_NOTIFY
+ data->consume_head = 0;
+ data->produce_tail = 0;
+#ifdef VSOCK_OPTIMIZATION_FLOW_CONTROL
+ data->notify_on_block = false;
+
+ if (PKT_FIELD(vsk, write_notify_min_window) < target + 1) {
+ ASSERT(target < vsk->consume_size);
+ PKT_FIELD(vsk, write_notify_min_window) = target + 1;
+ if (PKT_FIELD(vsk, write_notify_window) <
+ PKT_FIELD(vsk, write_notify_min_window)) {
+ /*
+ * If the current window is smaller than the new
+ * minimal window size, we need to reevaluate whether
+ * we need to notify the sender. If the number of ready
+ * bytes are smaller than the new window, we need to
+ * send a notification to the sender before we block.
+ */
+
+ PKT_FIELD(vsk, write_notify_window) =
+ PKT_FIELD(vsk, write_notify_min_window);
+ data->notify_on_block = true;
+ }
+ }
+#endif
+#endif
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_recv_pre_block --
+ *
+ * Called right before a socket is about to block with the socket lock held.
+ * The socket lock may have been released between the entry function and the
+ * preblock call.
+ *
+ * Note: This function may be called multiple times before the post block
+ * function is called.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_recv_pre_block(struct sock *sk,
+ size_t target,
+ vsock_vmci_recv_notify_data *data)
+{
+ int err = 0;
+
+ /* Notify our peer that we are waiting for data to read. */
+ if (!vsock_vmci_send_waiting_read(sk, target)) {
+ err = -EHOSTUNREACH;
+ return err;
+ }
+#ifdef VSOCK_OPTIMIZATION_FLOW_CONTROL
+ if (data->notify_on_block) {
+ err = vsock_vmci_send_read_notification(sk);
+ if (err < 0)
+ return err;
+
+ data->notify_on_block = false;
+ }
+#endif
+
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_recv_pre_dequeue --
+ *
+ * Called right before we dequeue / peek data from a socket.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_recv_pre_dequeue(struct sock *sk,
+ size_t target,
+ vsock_vmci_recv_notify_data *data)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ /*
+ * Now consume up to len bytes from the queue. Note that since we have
+ * the socket locked we should copy at least ready bytes.
+ */
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vmci_qpair_get_consume_indexes(vsk->qpair,
+ &data->produce_tail,
+ &data->consume_head);
+#endif
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_recv_post_dequeue --
+ *
+ * Called right after we dequeue / peek data from a socket.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_recv_post_dequeue(struct sock *sk,
+ size_t target,
+ ssize_t copied,
+ bool data_read,
+ vsock_vmci_recv_notify_data *data)
+{
+ vsock_vmci_sock *vsk;
+ int err;
+
+ vsk = vsock_sk(sk);
+ err = 0;
+
+ if (data_read) {
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ /*
+ * Detect a wrap-around to maintain queue generation. Note
+ * that this is safe since we hold the socket lock across the
+ * two queue pair operations.
+ */
+ if (copied >= vsk->consume_size - data->consume_head)
+ PKT_FIELD(vsk, consume_q_generation)++;
+
+#endif
+
+ err = vsock_vmci_send_read_notification(sk);
+ if (err < 0)
+ return err;
+
+ }
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_send_init --
+ *
+ * Called at the start of a stream send call with the socket lock held.
+ *
+ * Results: 0 on success. A negative error code on failure.
+ *
+ * Side effects:
+ */
+
+static int
+vsock_vmci_notify_pkt_send_init(struct sock *sk,
+ vsock_vmci_send_notify_data *data)
+{
+#ifdef VSOCK_OPTIMIZATION_WAITING_NOTIFY
+ data->consume_head = 0;
+ data->produce_tail = 0;
+#endif
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_send_pre_block --
+ *
+ * Called right before a socket is about to block with the socket lock held.
+ * The socket lock may have been released between the entry function and the
+ * preblock call.
+ *
+ * Note: This function may be called multiple times before the post block
+ * function is called.
+ *
+ * Results. 0 on success. A negative error code on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_send_pre_block(struct sock *sk,
+ vsock_vmci_send_notify_data *data)
+{
+ /* Notify our peer that we are waiting for room to write. */
+ if (!vsock_vmci_send_waiting_write(sk, 1))
+ return -EHOSTUNREACH;
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notifySendPreEnqueue --
+ *
+ * Called right before we Enqueue to a socket.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_send_pre_enqueue(struct sock *sk,
+ vsock_vmci_send_notify_data *data)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ vmci_qpair_get_produce_indexes(vsk->qpair,
+ &data->produce_tail,
+ &data->consume_head);
+#endif
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notifySendPostEnqueue --
+ *
+ * Called right after we enqueue data to a socket.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_send_post_enqueue(struct sock *sk,
+ ssize_t written,
+ vsock_vmci_send_notify_data *data)
+{
+ int err = 0;
+ vsock_vmci_sock *vsk;
+ bool sent_wrote = false;
+ int retries = 0;
+
+ vsk = vsock_sk(sk);
+
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ /*
+ * Detect a wrap-around to maintain queue generation. Note that this
+ * is safe since we hold the socket lock across the two queue pair
+ * operations.
+ */
+ if (written >= vsk->produce_size - data->produce_tail)
+ PKT_FIELD(vsk, produce_q_generation)++;
+
+#endif
+
+ if (vsock_vmci_notify_waiting_read(vsk)) {
+ /*
+ * Notify the peer that we have written, retrying the send on
+ * failure up to our maximum value. See the XXX comment for the
+ * corresponding piece of code in StreamRecvmsg() for potential
+ * improvements.
+ */
+ while (!(vsk->peer_shutdown & RCV_SHUTDOWN) &&
+ !sent_wrote && retries < VSOCK_MAX_DGRAM_RESENDS) {
+ err = VSOCK_SEND_WROTE(sk);
+ if (err >= 0)
+ sent_wrote = true;
+
+ retries++;
+ }
+
+ if (retries >= VSOCK_MAX_DGRAM_RESENDS) {
+ printk
+ (" %p unable to send wrote notify to peer.\n",
+ sk);
+ return err;
+ } else {
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+ PKT_FIELD(vsk, peer_waiting_read) = false;
+#endif
+ }
+ }
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_handle_pkt
+ *
+ * Called when a notify packet is recieved for a socket in the connected state.
+ * Note this might be called from a bottom half.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void
+vsock_vmci_notify_pkt_handle_pkt(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst,
+ struct sockaddr_vm *src, bool *pkt_processed)
+{
+ bool processed = false;
+
+ switch (pkt->type) {
+ case VSOCK_PACKET_TYPE_WROTE:
+ vsock_vmci_handle_wrote(sk, pkt, bottom_half, dst, src);
+ processed = true;
+ break;
+ case VSOCK_PACKET_TYPE_READ:
+ vsock_vmci_handle_read(sk, pkt, bottom_half, dst, src);
+ processed = true;
+ break;
+ case VSOCK_PACKET_TYPE_WAITING_WRITE:
+ vsock_vmci_handle_waiting_write(sk, pkt, bottom_half, dst, src);
+ processed = true;
+ break;
+
+ case VSOCK_PACKET_TYPE_WAITING_READ:
+ vsock_vmci_handle_waiting_read(sk, pkt, bottom_half, dst, src);
+ processed = true;
+ break;
+ }
+
+ if (pkt_processed)
+ *pkt_processed = processed;
+
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_process_request
+ *
+ * Called near the end of process request.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_process_request(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, write_notify_window) = vsk->consume_size;
+ if (vsk->consume_size < PKT_FIELD(vsk, write_notify_min_window))
+ PKT_FIELD(vsk, write_notify_min_window) = vsk->consume_size;
+
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_process_negotiate
+ *
+ * Called near the end of process negotiate.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_process_negotiate(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, write_notify_window) = vsk->consume_size;
+ if (vsk->consume_size < PKT_FIELD(vsk, write_notify_min_window))
+ PKT_FIELD(vsk, write_notify_min_window) = vsk->consume_size;
+
+}
+
+/* Socket control packet based operations. */
+vsock_vmci_notify_ops vsock_vmci_notify_pkt_ops = {
+ vsock_vmci_notify_pkt_socket_init,
+ vsock_vmci_notify_pkt_socket_destruct,
+ vsock_vmci_notify_pkt_poll_in,
+ vsock_vmci_notify_pkt_poll_out,
+ vsock_vmci_notify_pkt_handle_pkt,
+ vsock_vmci_notify_pkt_recv_init,
+ vsock_vmci_notify_pkt_recv_pre_block,
+ vsock_vmci_notify_pkt_recv_pre_dequeue,
+ vsock_vmci_notify_pkt_recv_post_dequeue,
+ vsock_vmci_notify_pkt_send_init,
+ vsock_vmci_notify_pkt_send_pre_block,
+ vsock_vmci_notify_pkt_send_pre_enqueue,
+ vsock_vmci_notify_pkt_send_post_enqueue,
+ vsock_vmci_notify_pkt_process_request,
+ vsock_vmci_notify_pkt_process_negotiate,
+};
diff --git a/net/vmw_vsock/notify.h b/net/vmw_vsock/notify.h
new file mode 100644
index 0000000..fe39442
--- /dev/null
+++ b/net/vmw_vsock/notify.h
@@ -0,0 +1,130 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2009-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * notify.h --
+ *
+ * Notify functions for Linux VSocket module.
+ */
+
+#ifndef __NOTIFY_H__
+#define __NOTIFY_H__
+
+#include <linux/types.h>
+
+#include "vsock_common.h"
+#include "vsock_packet.h"
+
+/* Comment this out to compare with old protocol. */
+#define VSOCK_OPTIMIZATION_WAITING_NOTIFY 1
+#if defined(VSOCK_OPTIMIZATION_WAITING_NOTIFY)
+/* Comment this out to remove flow control for "new" protocol */
+#define VSOCK_OPTIMIZATION_FLOW_CONTROL 1
+#endif
+
+#define VSOCK_MAX_DGRAM_RESENDS 10
+
+#define NOTIFYCALLRET(vsk, rv, mod_fn, args...) \
+ do { \
+ if (vsk->notify_ops && \
+ vsk->notify_ops->mod_fn != NULL) \
+ rv = (vsk->notify_ops->mod_fn)(args); \
+ else \
+ rv = 0; \
+ \
+ } while (0)
+
+#define NOTIFYCALL(vsk, mod_fn, args...) \
+ do { \
+ if (vsk->notify_ops && \
+ vsk->notify_ops->mod_fn != NULL) \
+ (vsk->notify_ops->mod_fn)(args); \
+ \
+ } while (0)
+
+typedef struct vsock_vmci_notify_pkt {
+ u64 write_notify_window;
+ u64 write_notify_min_window;
+ bool peer_waiting_read;
+ bool peer_waiting_write;
+ bool peer_waiting_write_detected;
+ bool sent_waiting_read;
+ bool sent_waiting_write;
+ vsock_waiting_info peer_waiting_read_info;
+ vsock_waiting_info peer_waiting_write_info;
+ u64 produce_q_generation;
+ u64 consume_q_generation;
+} vsock_vmci_notify_pkt;
+
+typedef struct vsock_vmci_notify_pkt_q_state {
+ u64 write_notify_window;
+ u64 write_notify_min_window;
+ bool peer_waiting_write;
+ bool peer_waiting_write_detected;
+} vsock_vmci_notify_pkt_q_state;
+
+typedef union vsock_vmci_notify {
+ vsock_vmci_notify_pkt pkt;
+ vsock_vmci_notify_pkt_q_state pkt_q_state;
+} vsock_vmci_notify;
+
+typedef struct vsock_vmci_recv_notify_data {
+ u64 consume_head;
+ u64 produce_tail;
+ bool notify_on_block;
+} vsock_vmci_recv_notify_data;
+
+typedef struct vsock_vmci_send_notify_data {
+ u64 consume_head;
+ u64 produce_tail;
+} vsock_vmci_send_notify_data;
+
+/* Socket notification callbacks. */
+typedef struct vsock_vmci_notify_ops {
+ void (*socket_init) (struct sock *sk);
+ void (*socket_destruct) (struct sock *sk);
+ int (*poll_in) (struct sock *sk, size_t target,
+ bool *data_ready_now);
+ int (*poll_out) (struct sock *sk, size_t target,
+ bool *space_avail_now);
+ void (*handle_notify_pkt) (struct sock *sk, vsock_packet *pkt,
+ bool bottom_half, struct sockaddr_vm *dst,
+ struct sockaddr_vm *src,
+ bool *pkt_processed);
+ int (*recv_init) (struct sock *sk, size_t target,
+ vsock_vmci_recv_notify_data *data);
+ int (*recv_pre_block) (struct sock *sk, size_t target,
+ vsock_vmci_recv_notify_data *data);
+ int (*recv_pre_dequeue) (struct sock *sk, size_t target,
+ vsock_vmci_recv_notify_data *data);
+ int (*recv_post_dequeue) (struct sock *sk, size_t target,
+ ssize_t copied, bool data_read,
+ vsock_vmci_recv_notify_data *data);
+ int (*send_init) (struct sock *sk,
+ vsock_vmci_send_notify_data *data);
+ int (*send_pre_block) (struct sock *sk,
+ vsock_vmci_send_notify_data *data);
+ int (*send_pre_enqueue) (struct sock *sk,
+ vsock_vmci_send_notify_data *data);
+ int (*send_post_enqueue) (struct sock *sk, ssize_t written,
+ vsock_vmci_send_notify_data *data);
+ void (*process_request) (struct sock *sk);
+ void (*process_negotiate) (struct sock *sk);
+} vsock_vmci_notify_ops;
+
+extern vsock_vmci_notify_ops vsock_vmci_notify_pkt_ops;
+extern vsock_vmci_notify_ops vsock_vmci_notify_pkt_q_state_ops;
+
+#endif /* __NOTIFY_H__ */
VSOCK stats for VMCI Stream Sockets protocol.
Signed-off-by: George Zhang <[email protected]>
---
net/vmw_vsock/stats.c | 37 ++++++++
net/vmw_vsock/stats.h | 219 +++++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 256 insertions(+), 0 deletions(-)
create mode 100644 net/vmw_vsock/stats.c
create mode 100644 net/vmw_vsock/stats.h
diff --git a/net/vmw_vsock/stats.c b/net/vmw_vsock/stats.c
new file mode 100644
index 0000000..2d172d5
--- /dev/null
+++ b/net/vmw_vsock/stats.c
@@ -0,0 +1,37 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2009-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * stats.c --
+ *
+ * Linux stats for the VMCI Stream Sockets protocol.
+ */
+
+#include <linux/types.h>
+
+#include <linux/socket.h>
+#include <linux/stddef.h> /* for NULL */
+#include <net/sock.h>
+
+#include "af_vsock.h"
+#include "stats.h"
+
+#ifdef VSOCK_GATHER_STATISTICS
+u64 vsock_stats_ctl_pkt_count[VSOCK_PACKET_TYPE_MAX];
+u64 vsock_stats_consume_queue_hist[VSOCK_NUM_QUEUE_LEVEL_BUCKETS];
+u64 vsock_stats_produce_queue_hist[VSOCK_NUM_QUEUE_LEVEL_BUCKETS];
+atomic64_t vsock_stats_consume_total;
+atomic64_t vsock_stats_produce_total;
+#endif
diff --git a/net/vmw_vsock/stats.h b/net/vmw_vsock/stats.h
new file mode 100644
index 0000000..2f9d44c
--- /dev/null
+++ b/net/vmw_vsock/stats.h
@@ -0,0 +1,219 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2009-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * stats.h --
+ *
+ * Stats functions for Linux vsock module.
+ */
+
+#ifndef __STATS_H__
+#define __STATS_H__
+
+#include <linux/types.h>
+
+#include "vsock_common.h"
+#include "vsock_packet.h"
+
+/*
+ * Define VSOCK_GATHER_STATISTICS to turn on statistics gathering. Currently
+ * this consists of 3 types of stats: 1. The number of control datagram
+ * messages sent. 2. The level of queuepair fullness (in 10% buckets) whenever
+ * data is about to be enqueued or dequeued from the queuepair. 3. The total
+ * number of bytes enqueued/dequeued.
+ */
+
+#ifdef VSOCK_GATHER_STATISTICS
+
+#define VSOCK_NUM_QUEUE_LEVEL_BUCKETS 10
+extern u64 vsock_stats_ctl_pkt_count[VSOCK_PACKET_TYPE_MAX];
+extern u64 vsock_stats_consume_queue_hist[VSOCK_NUM_QUEUE_LEVEL_BUCKETS];
+extern u64 vsock_stats_produce_queue_hist[VSOCK_NUM_QUEUE_LEVEL_BUCKETS];
+extern atomic64_t vsock_stats_consume_total;
+extern atomic64_t vsock_stats_produce_total;
+
+#define VSOCK_STATS_STREAM_CONSUME_HIST(vsk) \
+ vsock_vmci_stats_update_queue_bucket_count((vsk)->qpair, \
+ (vsk)->consume_size, \
+ vmci_qpair_consume_buf_ready((vsk)->qpair), \
+ vsock_stats_consume_queue_hist)
+#define VSOCK_STATS_STREAM_PRODUCE_HIST(vsk) \
+ vsock_vmci_stats_update_queue_bucket_count((vsk)->qpair, \
+ (vsk)->produce_size, \
+ vmci_qpair_produce_buf_ready((vsk)->qpair), \
+ vsock_stats_produce_queue_hist)
+#define VSOCK_STATS_CTLPKT_LOG(pkt_type) \
+ do { \
+ ++vsock_stats_ctl_pkt_count[pkt_type]; \
+ } while (0)
+#define VSOCK_STATS_STREAM_CONSUME(bytes) \
+ atomic64_add(&vsock_stats_consume_total, bytes)
+#define VSOCK_STATS_STREAM_PRODUCE(bytes) \
+ atomic64_add(&vsock_stats_produce_total, bytes)
+#define VSOCK_STATS_CTLPKT_DUMP_ALL() vsock_vmci_stats_ctl_pkt_dump_all()
+#define VSOCK_STATS_HIST_DUMP_ALL() vsock_vmci_stats_hist_dump_all()
+#define VSOCK_STATS_TOTALS_DUMP_ALL() vsock_vmci_stats_totals_dump_all()
+#define VSOCK_STATS_RESET() vsock_vmci_stats_reset()
+
+/*
+ *
+ * vsock_vmci_stats_update_queue_bucket_count --
+ *
+ * Given a queue, determine how much data is enqueued and add that to the
+ * specified queue level statistic bucket.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static inline void
+vsock_vmci_stats_update_queue_bucket_count(vmci_qpair *qpair,
+ u64 queue_size,
+ u64 data_ready,
+ u64 queue_hist[])
+{
+ u64 bucket = 0;
+ u32 remainder = 0;
+
+ /*
+ * We can't do 64 / 64 = 64 bit divides on linux because it requires a
+ * libgcc which is not linked into the kernel module. Since this code
+ * is only used by developers we just limit the queue_size to be less
+ * than MAX_UINT for now.
+ */
+ ASSERT(queue_size <= MAX_UINT32);
+ Div643264(data_ready * 10, queue_size, &bucket, &remainder);
+ ASSERT(bucket < VSOCK_NUM_QUEUE_LEVEL_BUCKETS);
+ ++queue_hist[bucket];
+}
+
+/*
+ *
+ * vsock_vmci_stats_ctl_pkt_dump_all --
+ *
+ * Prints all stream control packet counts out to the console using the
+ * appropriate platform logging.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static inline void vsock_vmci_stats_ctl_pkt_dump_all(void)
+{
+ int index;
+
+ ASSERT_ON_COMPILE(VSOCK_PACKET_TYPE_MAX ==
+ ARRAY_SIZE(vsock_stats_ctl_pkt_count));
+
+ for (index = 0; index < ARRAY_SIZE(vsock_stats_ctl_pkt_count);
+ index++) {
+ pr_info("Control packet: Type = %u, Count = %" FMT64
+ "u\n", index, vsock_stats_ctl_pkt_count[index]);
+ }
+}
+
+/*
+ *
+ * vsock_vmci_stats_hist_dump_all --
+ *
+ * Prints the produce and consume queue histograms to the console.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static inline void vsock_vmci_stats_hist_dump_all(void)
+{
+ int index;
+
+#define VSOCK_DUMP_HIST(strname, name) do { \
+ for (index = 0; index < ARRAY_SIZE(name); index++) { \
+ printk(strname " Bucket count %u = %"FMT64"u\n", \
+ index, name[index]); \
+ } \
+ } while (0)
+
+ VSOCK_DUMP_HIST("Produce Queue", vsock_stats_produce_queue_hist);
+ VSOCK_DUMP_HIST("Consume Queue", vsock_stats_consume_queue_hist);
+
+#undef VSOCK_DUMP_HIST
+}
+
+/*
+ *
+ * vsock_vmci_stats_totals_dump_all --
+ *
+ * Prints the produce and consume totals.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static inline void vsock_vmci_stats_totals_dump_all(void)
+{
+ pr_info("Produced %" FMT64 "u total bytes\n",
+ atomic64_read(&vsock_stats_produce_total));
+ pr_info("Consumed %" FMT64 "u total bytes\n",
+ atomic64_read(&vsock_stats_consume_total));
+}
+
+/*
+ *
+ * vsock_vmci_stats_reset --
+ *
+ * Reset all VSock statistics.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static inline void vsock_vmci_stats_reset(void)
+{
+ int index;
+
+#define VSOCK_RESET_ARRAY(name) do { \
+ for (index = 0; index < ARRAY_SIZE(name); index++) { \
+ name[index] = 0; \
+ } \
+ } while (0)
+
+ VSOCK_RESET_ARRAY(vsock_stats_ctl_pkt_count);
+ VSOCK_RESET_ARRAY(vsock_stats_produce_queue_hist);
+ VSOCK_RESET_ARRAY(vsock_stats_consume_queue_hist);
+
+#undef VSOCK_RESET_ARRAY
+
+ atomic64_set(&vsock_stats_consume_total, 0);
+ atomic64_set(&vsock_stats_produce_total, 0);
+}
+
+#else
+#define VSOCK_STATS_STREAM_CONSUME_HIST(vsk)
+#define VSOCK_STATS_STREAM_PRODUCE_HIST(vsk)
+#define VSOCK_STATS_STREAM_PRODUCE(bytes)
+#define VSOCK_STATS_STREAM_CONSUME(bytes)
+#define VSOCK_STATS_CTLPKT_LOG(pkt_type)
+#define VSOCK_STATS_CTLPKT_DUMP_ALL()
+#define VSOCK_STATS_HIST_DUMP_ALL()
+#define VSOCK_STATS_TOTALS_DUMP_ALL()
+#define VSOCK_STATS_RESET()
+#endif
+
+#endif
VSOCK utility functions for Linux VSocket module.
Signed-off-by: George Zhang <[email protected]>
---
net/vmw_vsock/util.c | 626 ++++++++++++++++++++++++++++++++++++++++++++++++++
net/vmw_vsock/util.h | 312 +++++++++++++++++++++++++
2 files changed, 938 insertions(+), 0 deletions(-)
create mode 100644 net/vmw_vsock/util.c
create mode 100644 net/vmw_vsock/util.h
diff --git a/net/vmw_vsock/util.c b/net/vmw_vsock/util.c
new file mode 100644
index 0000000..1c42b23
--- /dev/null
+++ b/net/vmw_vsock/util.c
@@ -0,0 +1,626 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * util.c --
+ *
+ * Utility functions for Linux VSocket module.
+ */
+
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/socket.h>
+#include <linux/stddef.h> /* for NULL */
+#include <net/sock.h>
+
+#include "af_vsock.h"
+#include "util.h"
+
+struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
+struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
+
+DEFINE_SPINLOCK(vsock_table_lock);
+
+/*
+ *
+ * vsock_vmci_log_pkt --
+ *
+ * Logs the provided packet.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+void vsock_vmci_log_pkt(char const *function, u32 line, vsock_packet *pkt)
+{
+ char buf[256];
+ char *cur = buf;
+ int left = sizeof buf;
+ int written = 0;
+ char *type_strings[] = {
+ [VSOCK_PACKET_TYPE_INVALID] = "INVALID",
+ [VSOCK_PACKET_TYPE_REQUEST] = "REQUEST",
+ [VSOCK_PACKET_TYPE_NEGOTIATE] = "NEGOTIATE",
+ [VSOCK_PACKET_TYPE_OFFER] = "OFFER",
+ [VSOCK_PACKET_TYPE_ATTACH] = "ATTACH",
+ [VSOCK_PACKET_TYPE_WROTE] = "WROTE",
+ [VSOCK_PACKET_TYPE_READ] = "READ",
+ [VSOCK_PACKET_TYPE_RST] = "RST",
+ [VSOCK_PACKET_TYPE_SHUTDOWN] = "SHUTDOWN",
+ [VSOCK_PACKET_TYPE_WAITING_WRITE] = "WAITING_WRITE",
+ [VSOCK_PACKET_TYPE_WAITING_READ] = "WAITING_READ",
+ [VSOCK_PACKET_TYPE_REQUEST2] = "REQUEST2",
+ [VSOCK_PACKET_TYPE_NEGOTIATE2] = "NEGOTIATE2",
+ };
+
+ written = snprintf(cur, left, "PKT: %u:%u -> %u:%u",
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src),
+ pkt->src_port,
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.dst),
+ pkt->dst_port);
+ if (written >= left)
+ goto error;
+
+ left -= written;
+ cur += written;
+
+ switch (pkt->type) {
+ case VSOCK_PACKET_TYPE_REQUEST:
+ case VSOCK_PACKET_TYPE_NEGOTIATE:
+ written = snprintf(cur, left, ", %s, size = %" FMT64 "u",
+ type_strings[pkt->type], pkt->u.size);
+ break;
+
+ case VSOCK_PACKET_TYPE_OFFER:
+ case VSOCK_PACKET_TYPE_ATTACH:
+ written = snprintf(cur, left, ", %s, handle = %u:%u",
+ type_strings[pkt->type],
+ VMCI_HANDLE_TO_CONTEXT_ID(pkt->u.handle),
+ VMCI_HANDLE_TO_RESOURCE_ID(pkt->u.handle));
+ break;
+
+ case VSOCK_PACKET_TYPE_WROTE:
+ case VSOCK_PACKET_TYPE_READ:
+ case VSOCK_PACKET_TYPE_RST:
+ written = snprintf(cur, left, ", %s", type_strings[pkt->type]);
+ break;
+ case VSOCK_PACKET_TYPE_SHUTDOWN: {
+ bool recv;
+ bool send;
+
+ recv = pkt->u.mode & RCV_SHUTDOWN;
+ send = pkt->u.mode & SEND_SHUTDOWN;
+ written = snprintf(cur, left, ", %s, mode = %c%c",
+ type_strings[pkt->type],
+ recv ? 'R' : ' ', send ? 'S' : ' ');
+ }
+ break;
+
+ case VSOCK_PACKET_TYPE_WAITING_WRITE:
+ case VSOCK_PACKET_TYPE_WAITING_READ:
+ written = snprintf(cur, left, ", %s, generation = %" FMT64 "u, "
+ "offset = %" FMT64 "u",
+ type_strings[pkt->type],
+ pkt->u.wait.generation, pkt->u.wait.offset);
+
+ break;
+
+ case VSOCK_PACKET_TYPE_REQUEST2:
+ case VSOCK_PACKET_TYPE_NEGOTIATE2:
+ written = snprintf(cur, left, ", %s, size = %" FMT64 "u, "
+ "proto = %u",
+ type_strings[pkt->type], pkt->u.size,
+ pkt->proto);
+ break;
+
+ default:
+ written = snprintf(cur, left, ", unrecognized type");
+ }
+
+ if (written >= left)
+ goto error;
+
+ left -= written;
+ cur += written;
+
+ written = snprintf(cur, left, " [%s:%u]\n", function, line);
+ if (written >= left)
+ goto error;
+
+ return;
+
+error:
+ pr_err("could not log packet\n");
+}
+
+/*
+ *
+ * vsock_vmci_init_tables --
+ *
+ * Initializes the tables used for socket lookup.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+void vsock_vmci_init_tables(void)
+{
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
+ INIT_LIST_HEAD(&vsock_bind_table[i]);
+
+ for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
+ INIT_LIST_HEAD(&vsock_connected_table[i]);
+}
+
+/*
+ *
+ * __vsock_vmci_insert_bound --
+ *
+ * Inserts socket into the bound table.
+ *
+ * Note that this assumes any necessary locks are held.
+ *
+ * Results: None.
+ *
+ * Side effects: The reference count for sk is incremented.
+ */
+
+void __vsock_vmci_insert_bound(struct list_head *list, struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ sock_hold(sk);
+ list_add(&vsk->bound_table, list);
+}
+
+/*
+ *
+ * __vsock_vmci_insert_connected --
+ *
+ * Inserts socket into the connected table.
+ *
+ * Note that this assumes any necessary locks are held.
+ *
+ * Results: None.
+ *
+ * Side effects: The reference count for sk is incremented.
+ */
+
+void __vsock_vmci_insert_connected(struct list_head *list, struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ sock_hold(sk);
+ list_add(&vsk->connected_table, list);
+}
+
+/*
+ *
+ * __vsock_vmci_remove_bound --
+ *
+ * Removes socket from the bound table.
+ *
+ * Note that this assumes any necessary locks are held.
+ *
+ * Results: None.
+ *
+ * Side effects: The reference count for sk is decremented.
+ */
+
+void __vsock_vmci_remove_bound(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+
+ ASSERT(__vsock_vmci_in_bound_table(sk));
+
+ vsk = vsock_sk(sk);
+
+ list_del_init(&vsk->bound_table);
+ sock_put(sk);
+}
+
+/*
+ *
+ * __vsock_vmci_remove_connected --
+ *
+ * Removes socket from the connected table.
+ *
+ * Note that this assumes any necessary locks are held.
+ *
+ * Results: None.
+ *
+ * Side effects: The reference count for sk is decremented.
+ */
+
+void __vsock_vmci_remove_connected(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+
+ ASSERT(__vsock_vmci_in_connected_table(sk));
+
+ vsk = vsock_sk(sk);
+
+ list_del_init(&vsk->connected_table);
+ sock_put(sk);
+}
+
+/*
+ *
+ * __vsock_vmci_find_bound_socket --
+ *
+ * Finds the socket corresponding to the provided address in the bound sockets
+ * hash table.
+ *
+ * Note that this assumes any necessary locks are held.
+ *
+ * Results: The sock structure if found, NULL if not found.
+ *
+ * Side effects: None.
+ */
+
+struct sock *__vsock_vmci_find_bound_socket(struct sockaddr_vm *addr)
+{
+ vsock_vmci_sock *vsk;
+
+ list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) {
+ if (vsock_addr_equals_addr_any(addr, &vsk->local_addr)) {
+ return sk_vsock(vsk);
+ }
+ }
+
+ return NULL;
+}
+
+/*
+ *
+ * __vsock_vmci_find_connected_socket --
+ *
+ * Finds the socket corresponding to the provided addresses in the connected
+ * sockets hash table.
+ *
+ * Note that this assumes any necessary locks are held.
+ *
+ * Results: The sock structure if found, NULL if not found.
+ *
+ * Side effects: None.
+ */
+
+struct sock *__vsock_vmci_find_connected_socket(struct sockaddr_vm *src,
+ struct sockaddr_vm *dst)
+{
+ vsock_vmci_sock *vsk;
+
+ list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
+ connected_table) {
+ if (vsock_addr_equals_addr(src, &vsk->remote_addr)
+ && vsock_addr_equals_addr(dst, &vsk->local_addr)) {
+ return sk_vsock(vsk);
+ }
+ }
+
+ return NULL;
+}
+
+/*
+ *
+ * __vsock_vmci_in_bound_table --
+ *
+ * Determines whether the provided socket is in the bound table.
+ *
+ * Results: TRUE is socket is in bound table, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool __vsock_vmci_in_bound_table(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ return !list_empty(&vsk->bound_table);
+}
+
+/*
+ *
+ * __vsock_vmci_in_connected_table --
+ *
+ * Determines whether the provided socket is in the connected table.
+ *
+ * Results: TRUE is socket is in connected table, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool __vsock_vmci_in_connected_table(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ return !list_empty(&vsk->connected_table);
+}
+
+/*
+ *
+ * vsock_vmci_get_pending --
+ *
+ * Retrieves a pending connection that matches the addresses specified in the
+ * provided packet.
+ *
+ * Assumes the socket lock is held for listener.
+ *
+ * Results: Socket of the pending connection on success, NULL if not found.
+ *
+ * Side effects: A reference is held on the socket until the release function
+ * is called.
+ */
+
+struct sock *vsock_vmci_get_pending(struct sock *listener, vsock_packet *pkt)
+{
+ vsock_vmci_sock *vlistener;
+ vsock_vmci_sock *vpending;
+ struct sock *pending;
+
+ vlistener = vsock_sk(listener);
+
+ list_for_each_entry(vpending, &vlistener->pending_links,
+ pending_links) {
+ struct sockaddr_vm src;
+ struct sockaddr_vm dst;
+
+ vsock_addr_init(&src, VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src),
+ pkt->src_port);
+ vsock_addr_init(&dst, VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.dst),
+ pkt->dst_port);
+
+ if (vsock_addr_equals_addr(&src, &vpending->remote_addr) &&
+ vsock_addr_equals_addr(&dst, &vpending->local_addr)) {
+ pending = sk_vsock(vpending);
+ sock_hold(pending);
+ goto found;
+ }
+ }
+
+ pending = NULL;
+found:
+ return pending;
+
+}
+
+/*
+ *
+ * vsock_vmci_release_pending --
+ *
+ * Releases the reference on a socket previously obtained by a call to
+ * vsock_vmci_get_pending().
+ *
+ * Results: None.
+ *
+ * Side effects: The socket may be freed if this was the last reference.
+ */
+
+void vsock_vmci_release_pending(struct sock *pending)
+{
+ sock_put(pending);
+}
+
+/*
+ *
+ * vsock_vmci_add_pending --
+ *
+ * Adds a pending connection on listener's pending list.
+ *
+ * Assumes the socket lock is held for listener. Assumes the socket lock is
+ * held for pending.
+ *
+ * Results: None.
+ *
+ * Side effects: The reference count of the sockets is incremented.
+ */
+
+void vsock_vmci_add_pending(struct sock *listener, struct sock *pending)
+{
+ vsock_vmci_sock *vlistener;
+ vsock_vmci_sock *vpending;
+
+ vlistener = vsock_sk(listener);
+ vpending = vsock_sk(pending);
+
+ sock_hold(pending);
+ sock_hold(listener);
+ list_add_tail(&vpending->pending_links, &vlistener->pending_links);
+}
+
+/*
+ *
+ * vsock_vmci_remove_pending --
+ *
+ * Removes a pending connection from the listener's pending list.
+ *
+ * Assumes the socket lock is held for listener. Assumes the socket lock is
+ * held for pending.
+ *
+ * Results: None.
+ *
+ * Side effects: The reference count of the sockets is decremented.
+ */
+
+void vsock_vmci_remove_pending(struct sock *listener, struct sock *pending)
+{
+ vsock_vmci_sock *vpending = vsock_sk(pending);
+
+ list_del_init(&vpending->pending_links);
+ sock_put(listener);
+ sock_put(pending);
+}
+
+/*
+ *
+ * vsock_vmci_enqueue_accept --
+ *
+ * Enqueues the connected socket on the listening socket's accepting queue.
+ *
+ * Assumes the socket lock is held for listener. Assumes the socket lock is
+ * held for connected.
+ *
+ * Results: None.
+ *
+ * Side effects: The sockets' reference counts are incremented.
+ */
+
+void vsock_vmci_enqueue_accept(struct sock *listener, struct sock *connected)
+{
+ vsock_vmci_sock *vlistener;
+ vsock_vmci_sock *vconnected;
+
+ vlistener = vsock_sk(listener);
+ vconnected = vsock_sk(connected);
+
+ sock_hold(connected);
+ sock_hold(listener);
+ list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
+}
+
+/*
+ *
+ * vsock_vmci_dequeue_accept --
+ *
+ * Dequeues the next connected socket from the listening socket's accept queue.
+ *
+ * Assumes the socket lock is held for listener.
+ *
+ * Note that the caller must call sock_put() on the returned socket once it is
+ * done with the socket.
+ *
+ * Results: The next socket from the queue, or NULL if the queue is empty.
+ *
+ * Side effects: The reference count of the listener is decremented.
+ */
+
+struct sock *vsock_vmci_dequeue_accept(struct sock *listener)
+{
+ vsock_vmci_sock *vlistener;
+ vsock_vmci_sock *vconnected;
+
+ vlistener = vsock_sk(listener);
+
+ if (list_empty(&vlistener->accept_queue))
+ return NULL;
+
+ vconnected = list_entry(vlistener->accept_queue.next,
+ vsock_vmci_sock, accept_queue);
+ ASSERT(vconnected);
+
+ list_del_init(&vconnected->accept_queue);
+ sock_put(listener);
+ /*
+ * The caller will need a reference on the connected socket so we let
+ * it call sock_put().
+ */
+
+ ASSERT(sk_vsock(vconnected));
+ return sk_vsock(vconnected);
+}
+
+/*
+ *
+ * vsock_vmci_remove_accept --
+ *
+ * Removes a socket from the accept queue of a listening socket.
+ *
+ * Assumes the socket lock is held for listener. Assumes the socket lock is
+ * held for connected.
+ *
+ * Results: None.
+ *
+ * Side effects: The sockets' reference counts are decremented.
+ */
+
+void vsock_vmci_remove_accept(struct sock *listener, struct sock *connected)
+{
+ vsock_vmci_sock *vconnected;
+
+ if (!vsock_vmci_in_accept_queue(connected))
+ return;
+
+ vconnected = vsock_sk(connected);
+ ASSERT(vconnected->listener == listener);
+
+ list_del_init(&vconnected->accept_queue);
+ sock_put(listener);
+ sock_put(connected);
+}
+
+/*
+ *
+ * vsock_vmci_in_accept_queue --
+ *
+ * Determines whether a socket is on an accept queue.
+ *
+ * Assumes the socket lock is held for sk.
+ *
+ * Results: TRUE if the socket is in an accept queue, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_vmci_in_accept_queue(struct sock *sk)
+{
+ /*
+ * If our accept queue isn't empty, it means we're linked into some
+ * listener socket's accept queue.
+ */
+ return !vsock_vmci_is_accept_queue_empty(sk);
+}
+
+/*
+ *
+ * vsock_vmci_is_accept_queue_empty --
+ *
+ * Determines whether the provided socket's accept queue is empty.
+ *
+ * Assumes the socket lock is held for sk.
+ *
+ * Results: TRUE if the socket's accept queue is empty, FALSE otherwsise.
+ *
+ * Side effects: None.
+ *
+ */
+
+bool vsock_vmci_is_accept_queue_empty(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+ return list_empty(&vsk->accept_queue);
+}
+
+/*
+ *
+ * vsock_vmci_is_pending --
+ *
+ * Determines whether a socket is pending.
+ *
+ * Assumes the socket lock is held for sk.
+ *
+ * Results: TRUE if the socket is pending, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_vmci_is_pending(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+ return !list_empty(&vsk->pending_links);
+}
diff --git a/net/vmw_vsock/util.h b/net/vmw_vsock/util.h
new file mode 100644
index 0000000..50ed6c8
--- /dev/null
+++ b/net/vmw_vsock/util.h
@@ -0,0 +1,312 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * util.h --
+ *
+ * Utility functions for Linux VSocket module.
+ */
+
+#ifndef __UTIL_H__
+#define __UTIL_H__
+
+#include <linux/types.h>
+#include <linux/stddef.h> /* for NULL */
+#include <net/sock.h>
+#include <linux/spinlock.h>
+
+#include "vsock_common.h"
+#include "vsock_packet.h"
+
+/*
+ * Each bound VSocket is stored in the bind hash table and each connected
+ * VSocket is stored in the connected hash table.
+ *
+ * Unbound sockets are all put on the same list attached to the end of the hash
+ * table (vsock_unbound_sockets). Bound sockets are added to the hash table in
+ * the bucket that their local address hashes to (vsock_bound_sockets(addr)
+ * represents the list that addr hashes to).
+ *
+ * Specifically, we initialize the vsock_bind_table array to a size of
+ * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
+ * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
+ * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function
+ * mods with VSOCK_HASH_SIZE - 1 to ensure this.
+ */
+#define VSOCK_HASH_SIZE 251
+#define LAST_RESERVED_PORT 1023
+#define MAX_PORT_RETRIES 24
+
+extern struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
+extern struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
+
+extern spinlock_t vsock_table_lock;
+
+#define VSOCK_HASH(addr) ((addr)->svm_port % (VSOCK_HASH_SIZE - 1))
+#define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
+#define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE])
+
+/* XXX This can probably be implemented in a better way. */
+#define VSOCK_CONN_HASH(src, dst) \
+ (((src)->svm_cid ^ (dst)->svm_port) % (VSOCK_HASH_SIZE - 1))
+#define vsock_connected_sockets(src, dst) \
+ (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
+#define vsock_connected_sockets_vsk(vsk) \
+ vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
+
+/*
+ * Prototypes.
+ */
+
+void vsock_vmci_log_pkt(char const *function, u32 line, vsock_packet *pkt);
+
+void vsock_vmci_init_tables(void);
+void __vsock_vmci_insert_bound(struct list_head *list, struct sock *sk);
+void __vsock_vmci_insert_connected(struct list_head *list, struct sock *sk);
+void __vsock_vmci_remove_bound(struct sock *sk);
+void __vsock_vmci_remove_connected(struct sock *sk);
+struct sock *__vsock_vmci_find_bound_socket(struct sockaddr_vm *addr);
+struct sock *__vsock_vmci_find_connected_socket(struct sockaddr_vm *src,
+ struct sockaddr_vm *dst);
+bool __vsock_vmci_in_bound_table(struct sock *sk);
+bool __vsock_vmci_in_connected_table(struct sock *sk);
+
+struct sock *vsock_vmci_get_pending(struct sock *listener, vsock_packet *pkt);
+void vsock_vmci_release_pending(struct sock *pending);
+void vsock_vmci_add_pending(struct sock *listener, struct sock *pending);
+void vsock_vmci_remove_pending(struct sock *listener, struct sock *pending);
+void vsock_vmci_enqueue_accept(struct sock *listener, struct sock *connected);
+struct sock *vsock_vmci_dequeue_accept(struct sock *listener);
+void vsock_vmci_remove_accept(struct sock *listener, struct sock *connected);
+bool vsock_vmci_in_accept_queue(struct sock *sk);
+bool vsock_vmci_is_accept_queue_empty(struct sock *sk);
+bool vsock_vmci_is_pending(struct sock *sk);
+
+static inline void vsock_vmci_insert_bound(struct list_head *list,
+ struct sock *sk);
+static inline void vsock_vmci_insert_connected(struct list_head *list,
+ struct sock *sk);
+static inline void vsock_vmci_remove_bound(struct sock *sk);
+static inline void vsock_vmci_remove_connected(struct sock *sk);
+static inline struct sock *vsock_vmci_find_bound_socket(struct sockaddr_vm
+ *addr);
+static inline struct sock *vsock_vmci_find_connected_socket(struct sockaddr_vm
+ *src,
+ struct sockaddr_vm
+ *dst);
+static inline bool vsock_vmci_in_bound_table(struct sock *sk);
+static inline bool vsock_vmci_in_connected_table(struct sock *sk);
+
+/*
+ *
+ * vsock_vmci_insert_bound --
+ *
+ * Inserts socket into the bound table.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these may be called from tasklets.
+ *
+ * Results: None.
+ *
+ * Side effects: vsock_table_lock is acquired and released.
+ */
+
+static inline void
+vsock_vmci_insert_bound(struct list_head *list, struct sock *sk)
+{
+ spin_lock_bh(&vsock_table_lock);
+ __vsock_vmci_insert_bound(list, sk);
+ spin_unlock_bh(&vsock_table_lock);
+}
+
+/*
+ *
+ * vsock_vmci_insert_connected --
+ *
+ * Inserts socket into the connected table.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these may be called from tasklets.
+ *
+ * Results: None.
+ *
+ * Side effects: vsock_table_lock is acquired and released.
+ */
+
+static inline void
+vsock_vmci_insert_connected(struct list_head *list, struct sock *sk)
+{
+ spin_lock_bh(&vsock_table_lock);
+ __vsock_vmci_insert_connected(list, sk);
+ spin_unlock_bh(&vsock_table_lock);
+}
+
+/*
+ *
+ * vsock_vmci_remove_bound --
+ *
+ * Removes socket from the bound list.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these may be called from tasklets.
+ *
+ * Results: None.
+ *
+ * Side effects: vsock_table_lock is acquired and released.
+ */
+
+static inline void vsock_vmci_remove_bound(struct sock *sk)
+{
+ spin_lock_bh(&vsock_table_lock);
+ __vsock_vmci_remove_bound(sk);
+ spin_unlock_bh(&vsock_table_lock);
+}
+
+/*
+ *
+ * vsock_vmci_remove_connected --
+ *
+ * Removes socket from the connected list.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these may be called from tasklets.
+ *
+ * Results: None.
+ *
+ * Side effects: vsock_table_lock is acquired and released.
+ */
+
+static inline void vsock_vmci_remove_connected(struct sock *sk)
+{
+ spin_lock_bh(&vsock_table_lock);
+ __vsock_vmci_remove_connected(sk);
+ spin_unlock_bh(&vsock_table_lock);
+}
+
+/*
+ *
+ * vsock_vmci_find_bound_socket --
+ *
+ * Finds the socket corresponding to the provided address in the bound sockets
+ * hash table.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these are called from tasklets.
+ *
+ * Results: The sock structure if found, NULL on failure.
+ *
+ * Side effects: vsock_table_lock is acquired and released. The socket's
+ * reference count is increased.
+ */
+
+static inline struct sock *vsock_vmci_find_bound_socket(struct sockaddr_vm
+ *addr)
+{
+ struct sock *sk;
+
+ spin_lock_bh(&vsock_table_lock);
+ sk = __vsock_vmci_find_bound_socket(addr);
+ if (sk)
+ sock_hold(sk);
+
+ spin_unlock_bh(&vsock_table_lock);
+
+ return sk;
+}
+
+/*
+ *
+ * vsock_vmci_find_connected_socket --
+ *
+ * Finds the socket corresponding to the provided address in the connected
+ * sockets hash table.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these are called from tasklets.
+ *
+ * Results: The sock structure if found, NULL on failure.
+ *
+ * Side effects: vsock_table_lock is acquired and released. The socket's
+ * reference count is increased.
+ */
+
+static inline struct sock *vsock_vmci_find_connected_socket(struct sockaddr_vm
+ *src,
+ struct sockaddr_vm
+ *dst)
+{
+ struct sock *sk;
+
+ spin_lock_bh(&vsock_table_lock);
+ sk = __vsock_vmci_find_connected_socket(src, dst);
+ if (sk)
+ sock_hold(sk);
+
+ spin_unlock_bh(&vsock_table_lock);
+
+ return sk;
+}
+
+/*
+ *
+ * vsock_vmci_in_bound_table --
+ *
+ * Determines whether the provided socket is in the bound table.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these may be called from tasklets.
+ *
+ * Results: TRUE is socket is in bound table, FALSE otherwise.
+ *
+ * Side effects: vsock_table_lock is acquired and released.
+ */
+
+static inline bool vsock_vmci_in_bound_table(struct sock *sk)
+{
+ bool ret;
+
+ spin_lock_bh(&vsock_table_lock);
+ ret = __vsock_vmci_in_bound_table(sk);
+ spin_unlock_bh(&vsock_table_lock);
+
+ return ret;
+}
+
+/*
+ *
+ * vsock_vmci_in_connected_table --
+ *
+ * Determines whether the provided socket is in the connected table.
+ *
+ * Note that it is important to invoke the bottom-half versions of the spinlock
+ * functions since these may be called from tasklets.
+ *
+ * Results: TRUE is socket is in connected table, FALSE otherwise.
+ *
+ * Side effects: vsock_table_lock is acquired and released.
+ */
+
+static inline bool vsock_vmci_in_connected_table(struct sock *sk)
+{
+ bool ret;
+
+ spin_lock_bh(&vsock_table_lock);
+ ret = __vsock_vmci_in_connected_table(sk);
+ spin_unlock_bh(&vsock_table_lock);
+
+ return ret;
+}
+
+#endif /* __UTIL_H__ */
VSOCK header files, Makefiles and Kconfig systems for Linux VSocket module.
Signed-off-by: George Zhang <[email protected]>
---
include/linux/socket.h | 4
net/Kconfig | 1
net/Makefile | 1
net/vmw_vsock/Kconfig | 14 +
net/vmw_vsock/Makefile | 4
net/vmw_vsock/notify_qstate.c | 627 +++++++++++++++++++++++++++++++++++
net/vmw_vsock/vmci_sockets.h | 517 +++++++++++++++++++++++++++++
net/vmw_vsock/vmci_sockets_packet.h | 106 ++++++
net/vmw_vsock/vsock_common.h | 130 +++++++
net/vmw_vsock/vsock_packet.h | 131 +++++++
net/vmw_vsock/vsock_version.h | 29 ++
11 files changed, 1563 insertions(+), 1 deletions(-)
create mode 100644 net/vmw_vsock/Kconfig
create mode 100644 net/vmw_vsock/Makefile
create mode 100644 net/vmw_vsock/notify_qstate.c
create mode 100644 net/vmw_vsock/vmci_sockets.h
create mode 100644 net/vmw_vsock/vmci_sockets_packet.h
create mode 100644 net/vmw_vsock/vsock_common.h
create mode 100644 net/vmw_vsock/vsock_packet.h
create mode 100644 net/vmw_vsock/vsock_version.h
diff --git a/include/linux/socket.h b/include/linux/socket.h
index 25d6322..57bc85e 100644
--- a/include/linux/socket.h
+++ b/include/linux/socket.h
@@ -195,7 +195,8 @@ struct ucred {
#define AF_CAIF 37 /* CAIF sockets */
#define AF_ALG 38 /* Algorithm sockets */
#define AF_NFC 39 /* NFC sockets */
-#define AF_MAX 40 /* For now.. */
+#define AF_VSOCK 40 /* VMCI sockets */
+#define AF_MAX 41 /* For now.. */
/* Protocol families, same as address families. */
#define PF_UNSPEC AF_UNSPEC
@@ -238,6 +239,7 @@ struct ucred {
#define PF_CAIF AF_CAIF
#define PF_ALG AF_ALG
#define PF_NFC AF_NFC
+#define PF_VSOCK AF_VSOCK
#define PF_MAX AF_MAX
/* Maximum queue length specifiable by listen. */
diff --git a/net/Kconfig b/net/Kconfig
index 245831b..75b8d5e 100644
--- a/net/Kconfig
+++ b/net/Kconfig
@@ -216,6 +216,7 @@ source "net/dcb/Kconfig"
source "net/dns_resolver/Kconfig"
source "net/batman-adv/Kconfig"
source "net/openvswitch/Kconfig"
+source "net/vmw_vsock/Kconfig"
config RPS
boolean
diff --git a/net/Makefile b/net/Makefile
index 4f4ee08..cae59f4 100644
--- a/net/Makefile
+++ b/net/Makefile
@@ -70,3 +70,4 @@ obj-$(CONFIG_CEPH_LIB) += ceph/
obj-$(CONFIG_BATMAN_ADV) += batman-adv/
obj-$(CONFIG_NFC) += nfc/
obj-$(CONFIG_OPENVSWITCH) += openvswitch/
+obj-$(CONFIG_VMWARE_VSOCK) += vmw_vsock/
diff --git a/net/vmw_vsock/Kconfig b/net/vmw_vsock/Kconfig
new file mode 100644
index 0000000..95e2568
--- /dev/null
+++ b/net/vmw_vsock/Kconfig
@@ -0,0 +1,14 @@
+#
+# Vsock protocol
+#
+
+config VMWARE_VSOCK
+ tristate "Virtual Socket protocol"
+ depends on VMWARE_VMCI
+ help
+ Virtual Socket Protocol is a socket protocol similar to TCP/IP
+ allowing comunication between Virtual Machines and VMware
+ hypervisor.
+
+ To compile this driver as a module, choose M here: the module
+ will be called vsock. If unsure, say N.
diff --git a/net/vmw_vsock/Makefile b/net/vmw_vsock/Makefile
new file mode 100644
index 0000000..4e940fe
--- /dev/null
+++ b/net/vmw_vsock/Makefile
@@ -0,0 +1,4 @@
+obj-$(CONFIG_VMWARE_VSOCK) += vmw_vsock.o
+
+vmw_vsock-y += af_vsock.o notify.o notify_qstate.o stats.o util.o \
+ vsock_addr.o
diff --git a/net/vmw_vsock/notify_qstate.c b/net/vmw_vsock/notify_qstate.c
new file mode 100644
index 0000000..9cc4b58
--- /dev/null
+++ b/net/vmw_vsock/notify_qstate.c
@@ -0,0 +1,627 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2009-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * notifyQState.c --
+ *
+ * Linux control notifications based on Queuepair state for the VMCI Stream
+ * Sockets protocol.
+ */
+
+#include <linux/types.h>
+
+#include <linux/socket.h>
+
+#include <linux/stddef.h> /* for NULL */
+#include <net/sock.h>
+
+#include "notify.h"
+#include "af_vsock.h"
+
+#define PKT_FIELD(vsk, field_name) \
+ (vsk)->notify.pkt_q_state.field_name
+
+/*
+ *
+ * vsock_vmci_notify_waiting_write --
+ *
+ * Determines if the conditions have been met to notify a waiting writer.
+ *
+ * Results: true if a notification should be sent, false otherwise.
+ *
+ * Side effects: None.
+ */
+
+static bool vsock_vmci_notify_waiting_write(vsock_vmci_sock *vsk)
+{
+ bool retval;
+ u64 notify_limit;
+
+ if (!PKT_FIELD(vsk, peer_waiting_write))
+ return false;
+
+ /*
+ * When the sender blocks, we take that as a sign that the sender is
+ * faster than the receiver. To reduce the transmit rate of the sender,
+ * we delay the sending of the read notification by decreasing the
+ * write_notify_window. The notification is delayed until the number of
+ * bytes used in the queue drops below the write_notify_window.
+ */
+
+ if (!PKT_FIELD(vsk, peer_waiting_write_detected)) {
+ PKT_FIELD(vsk, peer_waiting_write_detected) = true;
+ if (PKT_FIELD(vsk, write_notify_window) < PAGE_SIZE) {
+ PKT_FIELD(vsk, write_notify_window) =
+ PKT_FIELD(vsk, write_notify_min_window);
+ } else {
+ PKT_FIELD(vsk, write_notify_window) -= PAGE_SIZE;
+ if (PKT_FIELD(vsk, write_notify_window) <
+ PKT_FIELD(vsk, write_notify_min_window))
+ PKT_FIELD(vsk, write_notify_window) =
+ PKT_FIELD(vsk, write_notify_min_window);
+
+ }
+ }
+ notify_limit = vsk->consume_size - PKT_FIELD(vsk, write_notify_window);
+
+ /*
+ * The notify_limit is used to delay notifications in the case where
+ * flow control is enabled. Below the test is expressed in terms of
+ * free space in the queue: if free_space > ConsumeSize -
+ * write_notify_window then notify An alternate way of expressing this
+ * is to rewrite the expression to use the data ready in the receive
+ * queue: if write_notify_window > bufferReady then notify as
+ * free_space == ConsumeSize - bufferReady.
+ */
+
+ retval = vmci_qpair_consume_free_space(vsk->qpair) > notify_limit;
+
+ if (retval) {
+ /*
+ * Once we notify the peer, we reset the detected flag so the
+ * next wait will again cause a decrease in the window size.
+ */
+
+ PKT_FIELD(vsk, peer_waiting_write_detected) = false;
+ }
+ return retval;
+}
+
+/*
+ *
+ * vsock_vmci_handle_read --
+ *
+ * Handles an incoming read message.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void
+vsock_vmci_handle_read(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst, struct sockaddr_vm *src)
+{
+
+ sk->sk_write_space(sk);
+}
+
+/*
+ *
+ * vsock_vmci_handle_wrote --
+ *
+ * Handles an incoming wrote message.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void
+vsock_vmci_handle_wrote(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst, struct sockaddr_vm *src)
+{
+ sk->sk_data_ready(sk, 0);
+}
+
+/*
+ *
+ * vsock_vmci_block_update_write_window --
+ *
+ * Updates the write window when we are blocking for data.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_block_update_write_window(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+
+ vsk = vsock_sk(sk);
+
+ if (PKT_FIELD(vsk, write_notify_window) < vsk->consume_size)
+ PKT_FIELD(vsk, write_notify_window) =
+ min(PKT_FIELD(vsk, write_notify_window) + PAGE_SIZE,
+ vsk->consume_size);
+
+}
+
+/*
+ *
+ * vsock_vmci_send_read_notification --
+ *
+ * Sends a read notification to this socket's peer.
+ *
+ * Results: >= 0 if the datagram is sent successfully, negative error value
+ * otherwise.
+ *
+ * Side effects: None.
+ */
+
+static int vsock_vmci_send_read_notification(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+ bool sent_read;
+ unsigned int retries;
+ int err;
+
+ vsk = vsock_sk(sk);
+ sent_read = false;
+ retries = 0;
+ err = 0;
+
+ if (vsock_vmci_notify_waiting_write(vsk)) {
+ /*
+ * Notify the peer that we have read, retrying the send on
+ * failure up to our maximum value. XXX For now we just log
+ * the failure, but later we should schedule a work item to
+ * handle the resend until it succeeds. That would require
+ * keeping track of work items in the vsk and cleaning them up
+ * upon socket close.
+ */
+ while (!(vsk->peer_shutdown & RCV_SHUTDOWN) &&
+ !sent_read && retries < VSOCK_MAX_DGRAM_RESENDS) {
+ err = VSOCK_SEND_READ(sk);
+ if (err >= 0)
+ sent_read = true;
+
+ retries++;
+ }
+
+ if (retries >= VSOCK_MAX_DGRAM_RESENDS && !sent_read)
+ printk
+ ("%p unable to send read notification to peer.\n",
+ sk);
+ else
+ PKT_FIELD(vsk, peer_waiting_write) = false;
+
+ }
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_socket_init --
+ *
+ * Function that is called after a socket is created and before any notify ops
+ * are used.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_socket_init(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+ vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, write_notify_window) = PAGE_SIZE;
+ PKT_FIELD(vsk, write_notify_min_window) = PAGE_SIZE;
+ PKT_FIELD(vsk, peer_waiting_write) = false;
+ PKT_FIELD(vsk, peer_waiting_write_detected) = false;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_socket_destruct --
+ *
+ * Function that is called when the socket is being released.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_socket_destruct(struct sock *sk)
+{
+ vsock_vmci_sock *vsk;
+ vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, write_notify_window) = PAGE_SIZE;
+ PKT_FIELD(vsk, write_notify_min_window) = PAGE_SIZE;
+ PKT_FIELD(vsk, peer_waiting_write) = false;
+ PKT_FIELD(vsk, peer_waiting_write_detected) = false;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_poll_in --
+ *
+ * Called by the poll function to figure out if there is data to read and to
+ * setup future notifications if needed. Only called on sockets that aren't
+ * shutdown for recv.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_poll_in(struct sock *sk,
+ size_t target, bool *data_ready_now)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ if (vsock_vmci_stream_has_data(vsk)) {
+ *data_ready_now = true;
+ } else {
+ /*
+ * We can't read right now because there is nothing in the
+ * queue. Ask for notifications when there is something to
+ * read.
+ */
+ if (sk->sk_state == SS_CONNECTED)
+ vsock_vmci_block_update_write_window(sk);
+
+ *data_ready_now = false;
+ }
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_poll_out
+ *
+ * Called by the poll function to figure out if there is space to write and to
+ * setup future notifications if needed. Only called on a connected socket that
+ * isn't shutdown for send.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_poll_out(struct sock *sk,
+ size_t target, bool *space_avail_now)
+{
+ s64 produce_q_free_space;
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ produce_q_free_space = vsock_vmci_stream_has_space(vsk);
+ if (produce_q_free_space > 0) {
+ *space_avail_now = true;
+ return 0;
+ } else if (produce_q_free_space == 0) {
+ /*
+ * This is a connected socket but we can't currently send data.
+ * Nothing else to do.
+ */
+ *space_avail_now = false;
+ }
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_recv_init --
+ *
+ * Called at the start of a stream recv call with the socket lock held.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_recv_init(struct sock *sk,
+ size_t target,
+ vsock_vmci_recv_notify_data *data)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ data->consume_head = 0;
+ data->produce_tail = 0;
+ data->notify_on_block = false;
+
+ if (PKT_FIELD(vsk, write_notify_min_window) < target + 1) {
+ ASSERT(target < vsk->consume_size);
+ PKT_FIELD(vsk, write_notify_min_window) = target + 1;
+ if (PKT_FIELD(vsk, write_notify_window) <
+ PKT_FIELD(vsk, write_notify_min_window)) {
+ /*
+ * If the current window is smaller than the new
+ * minimal window size, we need to reevaluate whether
+ * we need to notify the sender. If the number of ready
+ * bytes are smaller than the new window, we need to
+ * send a notification to the sender before we block.
+ */
+
+ PKT_FIELD(vsk, write_notify_window) =
+ PKT_FIELD(vsk, write_notify_min_window);
+ data->notify_on_block = true;
+ }
+ }
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_recv_pre_block --
+ *
+ * Called right before a socket is about to block with the socket lock held.
+ * The socket lock may have been released between the entry function and the
+ * preblock call.
+ *
+ * Note: This function may be called multiple times before the post block
+ * function is called.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_recv_pre_block(struct sock *sk,
+ size_t target,
+ vsock_vmci_recv_notify_data *data)
+{
+ int err = 0;
+
+ vsock_vmci_block_update_write_window(sk);
+
+ if (data->notify_on_block) {
+ err = vsock_vmci_send_read_notification(sk);
+ if (err < 0)
+ return err;
+
+ data->notify_on_block = false;
+ }
+
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_recv_post_dequeue --
+ *
+ * Called right after we dequeue / peek data from a socket.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_recv_post_dequeue(struct sock *sk,
+ size_t target,
+ ssize_t copied,
+ bool data_read,
+ vsock_vmci_recv_notify_data *data)
+{
+ vsock_vmci_sock *vsk;
+ int err;
+ bool was_full = false;
+ u64 free_space;
+
+ vsk = vsock_sk(sk);
+ err = 0;
+
+ if (data_read) {
+ smp_mb();
+
+ free_space = vmci_qpair_consume_free_space(vsk->qpair);
+ was_full = free_space == copied;
+
+ if (was_full)
+ PKT_FIELD(vsk, peer_waiting_write) = true;
+
+ err = vsock_vmci_send_read_notification(sk);
+ if (err < 0)
+ return err;
+
+ /* See the comment in vsock_vmci_notify_pkt_send_post_enqueue */
+ sk->sk_data_ready(sk, 0);
+ }
+
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_send_init --
+ *
+ * Called at the start of a stream send call with the socket lock held.
+ *
+ * Results: 0 on success. A negative error code on failure.
+ *
+ * Side effects:
+ */
+
+static int
+vsock_vmci_notify_pkt_send_init(struct sock *sk,
+ vsock_vmci_send_notify_data *data)
+{
+ data->consume_head = 0;
+ data->produce_tail = 0;
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_vmci_notifySendPostEnqueue --
+ *
+ * Called right after we enqueue data to a socket.
+ *
+ * Results: 0 on success. Negative error on failure.
+ *
+ * Side effects: None.
+ */
+
+static int
+vsock_vmci_notify_pkt_send_post_enqueue(struct sock *sk,
+ ssize_t written,
+ vsock_vmci_send_notify_data *data)
+{
+ int err = 0;
+ vsock_vmci_sock *vsk;
+ bool sent_wrote = false;
+ bool was_empty;
+ int retries = 0;
+
+ vsk = vsock_sk(sk);
+
+ smp_mb();
+
+ was_empty = (vmci_qpair_produce_buf_ready(vsk->qpair) == written);
+ if (was_empty) {
+ while (!(vsk->peer_shutdown & RCV_SHUTDOWN) &&
+ !sent_wrote && retries < VSOCK_MAX_DGRAM_RESENDS) {
+ err = VSOCK_SEND_WROTE(sk);
+ if (err >= 0)
+ sent_wrote = true;
+
+ retries++;
+ }
+ }
+
+ if (retries >= VSOCK_MAX_DGRAM_RESENDS && !sent_wrote) {
+ printk
+ ("%p unable to send wrote notification to peer.\n",
+ sk);
+ return err;
+ }
+
+ return err;
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_handle_pkt
+ *
+ * Called when a notify packet is recieved for a socket in the connected state.
+ * Note this might be called from a bottom half.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void
+vsock_vmci_notify_pkt_handle_pkt(struct sock *sk,
+ vsock_packet *pkt,
+ bool bottom_half,
+ struct sockaddr_vm *dst,
+ struct sockaddr_vm *src, bool *pkt_processed)
+{
+ bool processed = false;
+
+ switch (pkt->type) {
+ case VSOCK_PACKET_TYPE_WROTE:
+ vsock_vmci_handle_wrote(sk, pkt, bottom_half, dst, src);
+ processed = true;
+ break;
+ case VSOCK_PACKET_TYPE_READ:
+ vsock_vmci_handle_read(sk, pkt, bottom_half, dst, src);
+ processed = true;
+ break;
+ }
+
+ if (pkt_processed)
+ *pkt_processed = processed;
+
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_process_request
+ *
+ * Called near the end of process request.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_process_request(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, write_notify_window) = vsk->consume_size;
+ if (vsk->consume_size < PKT_FIELD(vsk, write_notify_min_window))
+ PKT_FIELD(vsk, write_notify_min_window) = vsk->consume_size;
+
+}
+
+/*
+ *
+ * vsock_vmci_notify_pkt_process_negotiate
+ *
+ * Called near the end of process negotiate.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static void vsock_vmci_notify_pkt_process_negotiate(struct sock *sk)
+{
+ vsock_vmci_sock *vsk = vsock_sk(sk);
+
+ PKT_FIELD(vsk, write_notify_window) = vsk->consume_size;
+ if (vsk->consume_size < PKT_FIELD(vsk, write_notify_min_window))
+ PKT_FIELD(vsk, write_notify_min_window) = vsk->consume_size;
+
+}
+
+/* Socket always on control packet based operations. */
+vsock_vmci_notify_ops vsock_vmci_notify_pkt_q_state_ops = {
+ vsock_vmci_notify_pkt_socket_init,
+ vsock_vmci_notify_pkt_socket_destruct,
+ vsock_vmci_notify_pkt_poll_in,
+ vsock_vmci_notify_pkt_poll_out,
+ vsock_vmci_notify_pkt_handle_pkt,
+ vsock_vmci_notify_pkt_recv_init,
+ vsock_vmci_notify_pkt_recv_pre_block,
+ NULL, /* recv_pre_dequeue */
+ vsock_vmci_notify_pkt_recv_post_dequeue,
+ vsock_vmci_notify_pkt_send_init,
+ NULL, /* send_pre_block */
+ NULL, /* send_pre_enqueue */
+ vsock_vmci_notify_pkt_send_post_enqueue,
+ vsock_vmci_notify_pkt_process_request,
+ vsock_vmci_notify_pkt_process_negotiate,
+};
diff --git a/net/vmw_vsock/vmci_sockets.h b/net/vmw_vsock/vmci_sockets.h
new file mode 100644
index 0000000..6e6fd98
--- /dev/null
+++ b/net/vmw_vsock/vmci_sockets.h
@@ -0,0 +1,517 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * vmci_sockets.h --
+ *
+ * VMCI sockets public constants and types.
+ */
+
+#ifndef _VMCI_SOCKETS_H_
+#define _VMCI_SOCKETS_H_
+
+#if !defined(__KERNEL__)
+#include <sys/socket.h>
+#endif
+
+/*
+ * \brief Option name for STREAM socket buffer size.
+ *
+ * Use as the option name in \c setsockopt(3) or \c getsockopt(3) to set or get
+ * an \c unsigned \c long \c long that specifies the size of the buffer
+ * underlying a vSockets STREAM socket.
+ *
+ * \note Value is clamped to the MIN and MAX.
+ *
+ * \see vmci_sock_get_af_value_fd() \see SO_VMCI_BUFFER_MIN_SIZE \see
+ * SO_VMCI_BUFFER_MAX_SIZE
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); unsigned long
+ * long val = 0x1000; int fd = socket(af, SOCK_STREAM, 0); setsockopt(fd, af,
+ * SO_VMCI_BUFFER_SIZE, &val, sizeof val); ... close(fd);
+ * vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define SO_VMCI_BUFFER_SIZE 0
+
+/*
+ * \brief Option name for STREAM socket minimum buffer size.
+ *
+ * Use as the option name in \c setsockopt(3) or \c getsockopt(3) to set or get
+ * an \c unsigned \c long \c long that specifies the minimum size allowed for
+ * the buffer underlying a vSockets STREAM socket.
+ *
+ * \see vmci_sock_get_af_value_fd() \see SO_VMCI_BUFFER_SIZE \see
+ * SO_VMCI_BUFFER_MAX_SIZE
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); unsigned long
+ * long val = 0x500; int fd = socket(af, SOCK_STREAM, 0); setsockopt(fd, af,
+ * SO_VMCI_BUFFER_MIN_SIZE, &val, sizeof val); ... close(fd);
+ * vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define SO_VMCI_BUFFER_MIN_SIZE 1
+
+/*
+ * \brief Option name for STREAM socket maximum buffer size.
+ *
+ * Use as the option name in \c setsockopt(3) or \c getsockopt(3) to set or get
+ * an unsigned long long that specifies the maximum size allowed for the buffer
+ * underlying a vSockets STREAM socket.
+ *
+ * \see vmci_sock_get_af_value_fd() \see SO_VMCI_BUFFER_SIZE \see
+ * SO_VMCI_BUFFER_MIN_SIZE
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); unsigned long
+ * long val = 0x4000; int fd = socket(af, SOCK_STREAM, 0); setsockopt(fd, af,
+ * SO_VMCI_BUFFER_MAX_SIZE, &val, sizeof val); ... close(fd);
+ * vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define SO_VMCI_BUFFER_MAX_SIZE 2
+
+/*
+ * \brief Option name for socket peer's host-specific VM ID.
+ *
+ * Use as the option name in \c getsockopt(3) to get a host-specific identifier
+ * for the peer endpoint's VM. The identifier is a signed integer.
+ *
+ * \note Only available for ESX (VMKernel/userworld) endpoints.
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); int id;
+ * socklen_t len = sizeof id; int fd = socket(af, SOCK_DGRAM, 0); getsockopt(fd,
+ * af, SO_VMCI_PEER_HOST_VM_ID, &id, &len); ... close(fd);
+ * vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define SO_VMCI_PEER_HOST_VM_ID 3
+
+/*
+ * \brief Option name for socket's service label.
+ *
+ * Use as the option name in \c setsockopt(3) or \c getsockopt(3) to set or get
+ * the service label for a socket. The service label is a C-style
+ * NUL-terminated string.
+ *
+ * \note Only available for ESX (VMkernel/userworld) endpoints.
+ */
+
+#define SO_VMCI_SERVICE_LABEL 4
+
+/*
+ * \brief Option name for determining if a socket is trusted.
+ *
+ * Use as the option name in \c getsockopt(3) to determine if a socket is
+ * trusted. The value is a signed integer.
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); int trusted;
+ * socklen_t len = sizeof trusted; int fd = socket(af, SOCK_DGRAM, 0);
+ * getsockopt(fd, af, SO_VMCI_TRUSTED, &trusted, &len); ... close(fd);
+ * vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define SO_VMCI_TRUSTED 5
+
+/*
+ * \brief Option name for STREAM socket connection timeout.
+ *
+ * Use as the option name in \c setsockopt(3) or \c getsockopt(3) to set or get
+ * the connection timeout for a STREAM socket. The value is platform
+ * dependent. On ESX, Linux and Mac OS, it is a \c struct \c timeval. On
+ * Windows, it is a \c DWORD.
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); struct
+ * timeval t = { 5, 100000 }; int fd = socket(af, SOCK_STREAM, 0);
+ * setsockopt(fd, af, SO_VMCI_CONNECT_TIMEOUT, &t, sizeof t); ... close(fd);
+ * vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define SO_VMCI_CONNECT_TIMEOUT 6
+
+/*
+ * \brief Option name for using non-blocking send/receive.
+ *
+ * Use as the option name for \c setsockopt(3) or \c getsockopt(3) to set or
+ * get the non-blocking transmit/receive flag for a STREAM socket. This flag
+ * determines whether \c send() and \c recv() can be called in non-blocking
+ * contexts for the given socket. The value is a signed integer.
+ *
+ * This option is only relevant to kernel endpoints, where descheduling the
+ * thread of execution is not allowed, for example, while holding a spinlock.
+ * It is not to be confused with conventional non-blocking socket operations.
+ *
+ * \note Only available for VMKernel endpoints.
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); int nonblock;
+ * socklen_t len = sizeof nonblock; int fd = socket(af, SOCK_STREAM, 0);
+ * getsockopt(fd, af, SO_VMCI_NONBLOCK_TXRX, &nonblock, &len); ... close(fd);
+ * vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define SO_VMCI_NONBLOCK_TXRX 7
+
+/*
+ * \brief The vSocket equivalent of INADDR_ANY.
+ *
+ * This works for the \c svm_cid field of sockaddr_vm and indicates the context
+ * ID of the current endpoint.
+ *
+ * \see sockaddr_vm
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); struct
+ * sockaddr_vm addr; int fd = socket(af, SOCK_DGRAM, 0); addr.svm_family = af;
+ * addr.svm_cid = VMADDR_CID_ANY; addr.svm_port = 2000; bind(fd, &addr, sizeof
+ * addr); ... close(fd); vmci_sock_release_af_value_fd(vmciFd); \endcode
+ */
+
+#define VMADDR_CID_ANY ((unsigned int)-1)
+
+/*
+ * \brief Bind to any available port.
+ *
+ * Works for the \c svm_port field of sockaddr_vm.
+ *
+ * \see sockaddr_vm
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); struct
+ * sockaddr_vm addr; int fd = socket(af, SOCK_DGRAM, 0); addr.svm_family = af;
+ * addr.svm_cid = VMADDR_CID_ANY; addr.svm_port = VMADDR_PORT_ANY; bind(fd,
+ * &addr, sizeof addr); ... close(fd); vmci_sock_release_af_value_fd(vmciFd);
+ * \endcode
+ */
+
+#define VMADDR_PORT_ANY ((unsigned int)-1)
+
+/*
+ * \brief Invalid vSockets version.
+ *
+ * \see VMCISock_Version()
+ */
+
+#define VMCI_SOCKETS_INVALID_VERSION ((unsigned int)-1)
+
+/*
+ * \brief The epoch (first) component of the vSockets version.
+ *
+ * A single byte representing the epoch component of the vSockets version.
+ *
+ * \see VMCISock_Version()
+ *
+ * An example is given below.
+ *
+ * \code unsigned int ver = VMCISock_Version(); unsigned char epoch =
+ * VMCI_SOCKETS_VERSION_EPOCH(ver); \endcode
+ */
+
+#define VMCI_SOCKETS_VERSION_EPOCH(_v) (((_v) & 0xFF000000) >> 24)
+
+/*
+ * \brief The major (second) component of the vSockets version.
+ *
+ * A single byte representing the major component of the vSockets version.
+ * Typically changes for every major release of a product.
+ *
+ * \see VMCISock_Version()
+ *
+ * An example is given below.
+ *
+ * \code unsigned int ver = VMCISock_Version(); unsigned char major =
+ * VMCI_SOCKETS_VERSION_MAJOR(ver); \endcode
+ */
+
+#define VMCI_SOCKETS_VERSION_MAJOR(_v) (((_v) & 0x00FF0000) >> 16)
+
+/*
+ * \brief The minor (third) component of the vSockets version.
+ *
+ * Two bytes representing the minor component of the vSockets version.
+ *
+ * \see VMCISock_Version()
+ *
+ * An example is given below.
+ *
+ * \code unsigned int ver = VMCISock_Version(); unsigned short minor =
+ * VMCI_SOCKETS_VERSION_MINOR(ver); \endcode
+ */
+
+#define VMCI_SOCKETS_VERSION_MINOR(_v) (((_v) & 0x0000FFFF))
+
+/** \cond PRIVATE */
+
+/** \endcond */
+
+/*
+ * \brief Address structure for vSockets.
+ *
+ * The address family should be set to whatever vmci_sock_get_af_value_fd()
+ * returns. The structure members should all align on their natural boundaries
+ * without resorting to compiler packing directives. The total size of this
+ * structure should be exactly the same as that of \c struct \c sockaddr.
+ *
+ * \see vmci_sock_get_af_value_fd()
+ */
+
+struct sockaddr_vm {
+
+ /** \brief Address family. \see vmci_sock_get_af_value_fd() */
+ sa_family_t svm_family;
+
+ /** \cond PRIVATE */
+ unsigned short svm_reserved1;
+ /** \endcond */
+
+ /** \brief Port. \see VMADDR_PORT_ANY */
+ unsigned int svm_port;
+
+ /** \brief Context ID. \see VMADDR_CID_ANY */
+ unsigned int svm_cid;
+
+ /** \cond PRIVATE */
+ unsigned char svm_zero[sizeof(struct sockaddr) -
+ sizeof(sa_family_t) -
+ sizeof(unsigned short) -
+ sizeof(unsigned int) - sizeof(unsigned int)];
+ /** \endcond */
+};
+
+#if defined(linux) && defined(__KERNEL__)
+int vmci_sock_get_local_c_id(void);
+#else
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <sys/ioctl.h>
+#include <unistd.h>
+
+#include <stdio.h>
+
+/** \cond PRIVATE */
+#define VMCI_SOCKETS_DEFAULT_DEVICE "/dev/vsock"
+#define VMCI_SOCKETS_CLASSIC_ESX_DEVICE "/vmfs/devices/char/vsock/vsock"
+#define VMCI_SOCKETS_VERSION 1972
+#define VMCI_SOCKETS_GET_AF_VALUE 1976
+#define VMCI_SOCKETS_GET_LOCAL_CID 1977
+/** \endcond */
+
+ /*
+ * VMCISock_Version
+ *
+ * \brief Retrieve the vSockets version.
+ *
+ * Returns the current version of vSockets. The version is a 32-bit
+ * unsigned integer that consist of three components: the epoch, the major
+ * version, and the minor version. Use the \c VMCI_SOCKETS_VERSION macros
+ * to extract the components.
+ *
+ * \see VMCI_SOCKETS_VERSION_EPOCH() \see VMCI_SOCKETS_VERSION_MAJOR() \see
+ * VMCI_SOCKETS_VERSION_MINOR()
+ *
+ * \retval VMCI_SOCKETS_INVALID_VERSION Not available. \retval other The
+ * current version.
+ *
+ * An example is given below.
+ *
+ * \code unsigned int ver = VMCISock_Version(); if (ver !=
+ * VMCI_SOCKETS_INVALID_VERSION) { printf("vSockets version=%d.%d.%d\n",
+ * VMCI_SOCKETS_VERSION_EPOCH(ver), VMCI_SOCKETS_VERSION_MAJOR(ver),
+ * VMCI_SOCKETS_VERSION_MINOR(ver)); } \endcode
+ */
+
+static inline unsigned int VMCISock_Version(void)
+{
+ int fd;
+ unsigned int version;
+
+ fd = open(VMCI_SOCKETS_DEFAULT_DEVICE, O_RDWR);
+ if (fd < 0) {
+ fd = open(VMCI_SOCKETS_CLASSIC_ESX_DEVICE, O_RDWR);
+ if (fd < 0)
+ return VMCI_SOCKETS_INVALID_VERSION;
+
+ }
+
+ if (ioctl(fd, VMCI_SOCKETS_VERSION, &version) < 0)
+ version = VMCI_SOCKETS_INVALID_VERSION;
+
+ close(fd);
+ return version;
+}
+
+ /*
+ * vmci_sock_get_af_value_fd
+ *
+ * \brief Retrieve the address family value for vSockets.
+ *
+ * Returns the value to be used for the VMCI Sockets address family. This
+ * value should be used as the domain argument to \c socket(2) (when you
+ * might otherwise use \c AF_INET). For VMCI Socket-specific options, this
+ * value should also be used for the level argument to \c setsockopt(2)
+ * (when you might otherwise use \c SOL_TCP).
+ *
+ * \see vmci_sock_release_af_value_fd() \see sockaddr_vm
+ *
+ * \param[out] out_fd File descriptor to the VMCI device. The address
+ * family value is valid until this descriptor is closed. This parameter is
+ * only valid if the return value is not -1. Call
+ * vmci_sock_release_af_value_fd() to close this descriptor.
+ *
+ * \retval -1 Not available. \retval other The address family value.
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); if (af !=
+ * -1) { int fd = socket(af, SOCK_STREAM, 0); ... close(fd); close(vmciFd); }
+ * \endcode
+ */
+
+static inline int vmci_sock_get_af_value_fd(int *out_fd)
+{
+ int fd;
+ int family;
+
+ fd = open(VMCI_SOCKETS_DEFAULT_DEVICE, O_RDWR);
+ if (fd < 0) {
+ fd = open(VMCI_SOCKETS_CLASSIC_ESX_DEVICE, O_RDWR);
+ if (fd < 0)
+ return -1;
+
+ }
+
+ if (ioctl(fd, VMCI_SOCKETS_GET_AF_VALUE, &family) < 0)
+ family = -1;
+
+ if (family < 0)
+ close(fd);
+ else if (out_fd)
+ *out_fd = fd;
+
+ return family;
+}
+
+ /** \cond PRIVATE */
+ /*
+ * vmci_sock_get_af_value
+ *
+ * \brief Retrieve the address family value for vSockets.
+ *
+ * Returns the value to be used for the VMCI Sockets address family. This
+ * value should be used as the domain argument to \c socket(2) (when you
+ * might otherwise use \c AF_INET). For VMCI Socket-specific options, this
+ * value should also be used for the level argument to \c setsockopt(2)
+ * (when you might otherwise use \c SOL_TCP).
+ *
+ * \note This function leaves its descriptor to the vsock device open so
+ * that the socket implementation knows that the socket family is still in
+ * use. This is done because the address family is registered with the
+ * kernel on-demand and a notification is needed to unregister the address
+ * family. Use of this function is thus discouraged; please use
+ * vmci_sock_get_af_value_fd() instead.
+ *
+ * \see vmci_sock_get_af_value_fd() \see sockaddr_vm
+ *
+ * \retval -1 Not available. \retval other The address family value.
+ *
+ * An example is given below.
+ *
+ * \code int af = vmci_sock_get_af_value(); if (af != -1) { int fd =
+ * socket(af, SOCK_STREAM, 0); ... close(fd); } \endcode
+ */
+
+static inline int vmci_sock_get_af_value(void)
+{
+ return vmci_sock_get_af_value_fd(NULL);
+}
+
+ /** \endcond PRIVATE */
+
+ /*
+ * vmci_sock_release_af_value_fd
+ *
+ * \brief Release the file descriptor obtained when retrieving the address
+ * family value.
+ *
+ * Use this to release the file descriptor obtained by calling
+ * vmci_sock_get_af_value_fd().
+ *
+ * \see vmci_sock_get_af_value_fd()
+ *
+ * \param[in] fd File descriptor to the VMCI device.
+ */
+
+static inline void vmci_sock_release_af_value_fd(int fd)
+{
+ if (fd >= 0)
+ close(fd);
+
+}
+
+ /*
+ * vmci_sock_get_local_c_id
+ *
+ * \brief Retrieve the current context ID.
+ *
+ * \see VMADDR_CID_ANY
+ *
+ * \retval VMADDR_CID_ANY Not available. \retval other The current context
+ * ID.
+ *
+ * An example is given below.
+ *
+ * \code int vmciFd; int af = vmci_sock_get_af_value_fd(&vmciFd); struct
+ * sockaddr_vm addr; addr.svm_family = af; addr.svm_cid =
+ * vmci_sock_get_local_c_id(); vmci_sock_release_af_value_fd(vmciFd);
+ * \endcode
+ */
+
+static inline unsigned int vmci_sock_get_local_c_id(void)
+{
+ int fd;
+ unsigned int context_id;
+
+ fd = open(VMCI_SOCKETS_DEFAULT_DEVICE, O_RDWR);
+ if (fd < 0) {
+ fd = open(VMCI_SOCKETS_CLASSIC_ESX_DEVICE, O_RDWR);
+ if (fd < 0)
+ return VMADDR_CID_ANY;
+
+ }
+
+ if (ioctl(fd, VMCI_SOCKETS_GET_LOCAL_CID, &context_id) < 0)
+ context_id = VMADDR_CID_ANY;
+
+ close(fd);
+ return context_id;
+}
+#endif
+
+#endif
diff --git a/net/vmw_vsock/vmci_sockets_packet.h b/net/vmw_vsock/vmci_sockets_packet.h
new file mode 100644
index 0000000..fc9f434
--- /dev/null
+++ b/net/vmw_vsock/vmci_sockets_packet.h
@@ -0,0 +1,106 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * vmci_sockets_packet.h --
+ *
+ * Definition of VMCI Sockets packet format, constants, and types.
+ */
+
+#ifndef _VMCI_SOCKETS_PACKET_H_
+#define _VMCI_SOCKETS_PACKET_H_
+
+#include <linux/vmw_vmci_defs.h>
+#include <linux/vmw_vmci_api.h>
+
+/*
+ * STREAM control packets.
+ */
+
+/* If the packet format changes in a release then this should change too. */
+#define VSOCK_PACKET_VERSION 1
+
+/* The resource ID on which control packets are sent. */
+#define VSOCK_PACKET_RID 1
+
+/*
+ * Assert that the given packet is valid. We check that the two original
+ * reserved fields equal zero because the version of the common code that
+ * shipped with ESX 4.0 and WS 6.5 did so and will return a RST packet if they
+ * aren't set that way. For newer packet types added after that release we
+ * don't do this.
+ */
+#define VSOCK_PACKET_ASSERT(_p) \
+ do { \
+ ASSERT((_p)->type < VSOCK_PACKET_TYPE_MAX); \
+ if ((_p)->type < VSOCK_PACKET_TYPE_REQUEST2) { \
+ ASSERT(0 == (_p)->proto); \
+ ASSERT(0 == (_p)->_reserved2); \
+ } \
+ } while (0)
+
+typedef enum vsock_packet_type {
+ VSOCK_PACKET_TYPE_INVALID = 0,
+ VSOCK_PACKET_TYPE_REQUEST,
+ VSOCK_PACKET_TYPE_NEGOTIATE,
+ VSOCK_PACKET_TYPE_OFFER,
+ VSOCK_PACKET_TYPE_ATTACH,
+ VSOCK_PACKET_TYPE_WROTE,
+ VSOCK_PACKET_TYPE_READ,
+ VSOCK_PACKET_TYPE_RST,
+ VSOCK_PACKET_TYPE_SHUTDOWN,
+ VSOCK_PACKET_TYPE_WAITING_WRITE,
+ VSOCK_PACKET_TYPE_WAITING_READ,
+ VSOCK_PACKET_TYPE_REQUEST2,
+ VSOCK_PACKET_TYPE_NEGOTIATE2,
+ VSOCK_PACKET_TYPE_MAX
+} vsock_packet_type;
+
+typedef u16 vsock_proto_version;
+#define VSOCK_PROTO_INVALID 0
+#define VSOCK_PROTO_PKT_ON_NOTIFY (1 << 0)
+
+#define VSOCK_PROTO_ALL_SUPPORTED (VSOCK_PROTO_PKT_ON_NOTIFY)
+
+typedef struct vsock_waiting_info {
+ u64 generation;
+ u64 offset;
+} vsock_waiting_info;
+
+/*
+ * Control packet type for STREAM sockets. DGRAMs have no control packets nor
+ * special packet header for data packets, they are just raw VMCI DGRAM
+ * messages. For STREAMs, control packets are sent over the control channel
+ * while data is written and read directly from queue pairs with no packet
+ * format.
+ */
+typedef struct vsock_packet {
+ struct vmci_datagram dg;
+ u8 version;
+ u8 type;
+ vsock_proto_version proto;
+
+ u32 src_port;
+ u32 dst_port;
+ u32 _reserved2;
+ union {
+ u64 size;
+ u64 mode;
+ struct vmci_handle handle;
+ vsock_waiting_info wait;
+ } u;
+} vsock_packet;
+
+#endif
diff --git a/net/vmw_vsock/vsock_common.h b/net/vmw_vsock/vsock_common.h
new file mode 100644
index 0000000..fd7f2cd
--- /dev/null
+++ b/net/vmw_vsock/vsock_common.h
@@ -0,0 +1,130 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * vsockCommon.h --
+ *
+ * VSockets common constants, types and functions.
+ */
+
+#ifndef _VSOCK_COMMON_H_
+#define _VSOCK_COMMON_H_
+
+/*
+ * vmci_sock_get_af_value_int is defined separately from vmci_sock_get_af_value
+ * because it is used in several different contexts. In particular it is called
+ * from vsock_addr.c which gets compiled into both our kernel modules as well
+ * as the user level vsock library. In the linux kernel we need different
+ * behavior than external kernel modules using VMCI Sockets api inside the
+ * kernel. FIXME
+ */
+
+#if defined __KERNEL__
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <asm/page.h>
+#else
+/* In userland, just use the normal exported userlevel api. */
+#define vmci_sock_get_af_value_int() vmci_sock_get_af_value()
+#endif
+
+#include <linux/vmw_vmci_defs.h>
+#include <linux/vmw_vmci_api.h>
+
+#include "vmci_sockets.h"
+#include "vsock_addr.h"
+
+#define ASSERT(cond) BUG_ON(!(cond))
+#define ASSERT_ON_COMPILE(cond) BUILD_BUG_ON(!(cond))
+
+#ifdef __x86_64__
+#define FMT64 "ll"
+#else
+#define FMT64 "L"
+#endif
+
+#ifndef PAGE_SHIFT
+#if defined __i386__
+#define PAGE_SHIFT 12
+#elif defined __arm__
+#define PAGE_SHIFT 12
+#else
+#error
+#endif
+#endif
+
+#ifndef PAGE_SIZE
+#define PAGE_SIZE (1<<PAGE_SHIFT)
+#endif
+
+#define MAX_UINT32 ((u32)0xffffffff)
+
+#ifndef ESYSNOTREADY
+#define ESYSNOTREADY EOPNOTSUPP
+#endif
+
+#define sockerr() errno
+#define sockerr2err(_e) (((_e) > 0) ? -(_e) : (_e))
+#define SS_LISTEN 255
+
+extern vmci_id vmci_get_context_id(void);
+
+/*
+ * Helper function to determine if the given handle points to the local context.
+ * Returns TRUE if the given handle is for the local context, FALSE otherwise.
+ */
+
+static inline bool vsock_vmci_is_local(struct vmci_handle handle)
+{
+ return vmci_get_context_id() == VMCI_HANDLE_TO_CONTEXT_ID(handle);
+}
+
+/*
+ * Helper function to convert from a VMCI error code to a VSock error code.
+ */
+
+static inline s32 vsock_vmci_error_to_vsock_error(s32 vmci_error)
+{
+ int err;
+
+ switch (vmci_error) {
+ case VMCI_ERROR_NO_MEM:
+ err = ENOMEM;
+ break;
+ case VMCI_ERROR_DUPLICATE_ENTRY:
+ case VMCI_ERROR_ALREADY_EXISTS:
+ err = EADDRINUSE;
+ break;
+ case VMCI_ERROR_NO_ACCESS:
+ err = EPERM;
+ break;
+ case VMCI_ERROR_NO_RESOURCES:
+ err = ENOBUFS;
+ break;
+ case VMCI_ERROR_INVALID_RESOURCE:
+ err = EHOSTUNREACH;
+ break;
+ case VMCI_ERROR_MODULE_NOT_LOADED:
+ err = ESYSNOTREADY;
+ break;
+ case VMCI_ERROR_INVALID_ARGS:
+ default:
+ err = EINVAL;
+ }
+
+ return sockerr2err(err);
+}
+
+#endif
diff --git a/net/vmw_vsock/vsock_packet.h b/net/vmw_vsock/vsock_packet.h
new file mode 100644
index 0000000..76bcac0
--- /dev/null
+++ b/net/vmw_vsock/vsock_packet.h
@@ -0,0 +1,131 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * vsockPacket.h --
+ *
+ * Packet constants, types and functions.
+ */
+
+#ifndef _VSOCK_PACKET_H_
+#define _VSOCK_PACKET_H_
+
+#include "vmci_sockets_packet.h"
+
+/*
+ *
+ * vsock_packet_init --
+ *
+ * Initialize the given packet. The packet version is set and the fields are
+ * filled out. Reserved fields are cleared.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static inline void
+vsock_packet_init(vsock_packet *pkt,
+ struct sockaddr_vm *src,
+ struct sockaddr_vm *dst,
+ u8 type,
+ u64 size,
+ u64 mode,
+ vsock_waiting_info *wait,
+ vsock_proto_version proto, struct vmci_handle handle)
+{
+ BUG_ON(!pkt);
+ BUG_ON(vsock_addr_validate(src) != 0);
+ BUG_ON(vsock_addr_validate(dst) != 0);
+
+ /*
+ * We register the stream control handler as an any cid handle so we
+ * must always send from a source address of VMADDR_CID_ANY
+ */
+ pkt->dg.src = VMCI_MAKE_HANDLE(VMADDR_CID_ANY, VSOCK_PACKET_RID);
+ pkt->dg.dst = VMCI_MAKE_HANDLE(dst->svm_cid, VSOCK_PACKET_RID);
+ pkt->dg.payload_size = sizeof *pkt - sizeof pkt->dg;
+ pkt->version = VSOCK_PACKET_VERSION;
+ pkt->type = type;
+ pkt->src_port = src->svm_port;
+ pkt->dst_port = dst->svm_port;
+ memset(&pkt->proto, 0, sizeof pkt->proto);
+ memset(&pkt->_reserved2, 0, sizeof pkt->_reserved2);
+
+ switch (pkt->type) {
+ case VSOCK_PACKET_TYPE_INVALID:
+ pkt->u.size = 0;
+ break;
+
+ case VSOCK_PACKET_TYPE_REQUEST:
+ case VSOCK_PACKET_TYPE_NEGOTIATE:
+ pkt->u.size = size;
+ break;
+
+ case VSOCK_PACKET_TYPE_OFFER:
+ case VSOCK_PACKET_TYPE_ATTACH:
+ pkt->u.handle = handle;
+ break;
+
+ case VSOCK_PACKET_TYPE_WROTE:
+ case VSOCK_PACKET_TYPE_READ:
+ case VSOCK_PACKET_TYPE_RST:
+ pkt->u.size = 0;
+ break;
+
+ case VSOCK_PACKET_TYPE_SHUTDOWN:
+ pkt->u.mode = mode;
+ break;
+
+ case VSOCK_PACKET_TYPE_WAITING_READ:
+ case VSOCK_PACKET_TYPE_WAITING_WRITE:
+ ASSERT(wait);
+ memcpy(&pkt->u.wait, wait, sizeof pkt->u.wait);
+ break;
+
+ case VSOCK_PACKET_TYPE_REQUEST2:
+ case VSOCK_PACKET_TYPE_NEGOTIATE2:
+ pkt->u.size = size;
+ pkt->proto = proto;
+ break;
+ }
+
+ VSOCK_PACKET_ASSERT(pkt);
+}
+
+/*
+ *
+ * vsock_packet_get_addresses --
+ *
+ * Get the local and remote addresses from the given packet.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+static inline void
+vsock_packet_get_addresses(vsock_packet *pkt,
+ struct sockaddr_vm *local,
+ struct sockaddr_vm *remote)
+{
+ VSOCK_PACKET_ASSERT(pkt);
+ vsock_addr_init(local, VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.dst),
+ pkt->dst_port);
+ vsock_addr_init(remote, VMCI_HANDLE_TO_CONTEXT_ID(pkt->dg.src),
+ pkt->src_port);
+}
+
+#endif
diff --git a/net/vmw_vsock/vsock_version.h b/net/vmw_vsock/vsock_version.h
new file mode 100644
index 0000000..9a6a63b
--- /dev/null
+++ b/net/vmw_vsock/vsock_version.h
@@ -0,0 +1,29 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2011-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * vsock_version.h --
+ *
+ * Version definitions for the Linux vsock driver.
+ */
+
+#ifndef _VSOCK_VERSION_H_
+#define _VSOCK_VERSION_H_
+
+#define VSOCK_DRIVER_VERSION 1.0.0.0-k
+#define VSOCK_DRIVER_VERSION_COMMAS 1,0,0,0
+#define VSOCK_DRIVER_VERSION_STRING "1.0.0.0-k"
+
+#endif /* _VSOCK_VERSION_H_ */
VSOCK linux address code implementation.
Signed-off-by: George Zhang <[email protected]>
---
net/vmw_vsock/vsock_addr.c | 264 ++++++++++++++++++++++++++++++++++++++++++++
net/vmw_vsock/vsock_addr.h | 40 +++++++
2 files changed, 304 insertions(+), 0 deletions(-)
create mode 100644 net/vmw_vsock/vsock_addr.c
create mode 100644 net/vmw_vsock/vsock_addr.h
diff --git a/net/vmw_vsock/vsock_addr.c b/net/vmw_vsock/vsock_addr.c
new file mode 100644
index 0000000..c93a174
--- /dev/null
+++ b/net/vmw_vsock/vsock_addr.c
@@ -0,0 +1,264 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * vsockAddr.c --
+ *
+ * VSockets address implementation.
+ */
+
+/*
+ * These includes come before vsockCommon.h to ensure that VMware's ASSERT
+ * macro is used instead of Linux's irda.h definition.
+ */
+#include <linux/types.h>
+#include <linux/socket.h>
+#include <linux/stddef.h> /* for NULL */
+#include <net/sock.h>
+
+#include "vsock_common.h"
+
+/*
+ *
+ * vsock_addr_init --
+ *
+ * Initialize the given address with the given context id and port. This will
+ * clear the address, set the correct family, and add the given values.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+void vsock_addr_init(struct sockaddr_vm *addr, u32 cid, u32 port)
+{
+ memset(addr, 0, sizeof *addr);
+
+ addr->svm_family = AF_VSOCK;
+ addr->svm_cid = cid;
+ addr->svm_port = port;
+
+ BUG_ON(vsock_addr_validate(addr) != 0);
+}
+
+/*
+ *
+ * vsock_addr_validate --
+ *
+ * Try to validate the given address. The address must not be null and must
+ * have the correct address family. Any reserved fields must be zero.
+ *
+ * Results: 0 on success, EFAULT if the address is null, EAFNOSUPPORT if the
+ * address is of the wrong family, and EINVAL if the reserved fields are not
+ * zero.
+ *
+ * Side effects: None.
+ */
+
+int vsock_addr_validate(const struct sockaddr_vm *addr)
+{
+ if (!addr)
+ return -EFAULT;
+
+ if (addr->svm_family != AF_VSOCK)
+ return -EAFNOSUPPORT;
+
+ if (addr->svm_zero[0] != 0)
+ return -EINVAL;
+
+ return 0;
+}
+
+/*
+ *
+ * vsock_addr_bound --
+ *
+ * Determines whether the provided address is bound.
+ *
+ * Results: TRUE if the address structure is bound, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_addr_bound(const struct sockaddr_vm *addr)
+{
+ BUG_ON(!addr);
+
+ return addr->svm_port != VMADDR_PORT_ANY;
+}
+
+/*
+ *
+ * vsock_addr_unbind --
+ *
+ * Unbind the given addresss.
+ *
+ * Results: None.
+ *
+ * Side effects: None.
+ */
+
+void vsock_addr_unbind(struct sockaddr_vm *addr)
+{
+ vsock_addr_init(addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
+}
+
+/*
+ *
+ * vsock_addr_equals_addr --
+ *
+ * Determine if the given addresses are equal.
+ *
+ * Results: TRUE if the addresses are equal, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_addr_equals_addr(const struct sockaddr_vm *addr,
+ const struct sockaddr_vm *other)
+{
+ BUG_ON(vsock_addr_validate(addr) != 0);
+ BUG_ON(vsock_addr_validate(other) != 0);
+
+ return addr->svm_cid == other->svm_cid &&
+ addr->svm_port == other->svm_port;
+}
+
+/*
+ *
+ * vsock_addr_equals_addr_any --
+ *
+ * Determine if the given addresses are equal. Will accept either an exact
+ * match or one where the rids match and that either the cids match or are set
+ * to VMADDR_CID_ANY.
+ *
+ * Results: TRUE if the addresses are equal, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_addr_equals_addr_any(const struct sockaddr_vm *addr,
+ const struct sockaddr_vm *other)
+{
+ BUG_ON(vsock_addr_validate(addr) != 0);
+ BUG_ON(vsock_addr_validate(other) != 0);
+
+ return (addr->svm_cid == VMADDR_CID_ANY ||
+ other->svm_cid == VMADDR_CID_ANY ||
+ addr->svm_cid == other->svm_cid) &&
+ addr->svm_port == other->svm_port;
+}
+
+/*
+ *
+ * vsock_addr_equals_handle_port --
+ *
+ * Determines if the given address matches the given handle and port.
+ *
+ * Results: TRUE if the address matches the handle and port, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_addr_equals_handle_port(const struct sockaddr_vm *addr,
+ struct vmci_handle handle, u32 port)
+{
+ BUG_ON(vsock_addr_validate(addr) != 0);
+
+ return addr->svm_cid == VMCI_HANDLE_TO_CONTEXT_ID(handle) &&
+ addr->svm_port == port;
+}
+
+/*
+ *
+ * vsock_addr_cast --
+ *
+ * Try to cast the given generic address to a VM address. The given length
+ * must match that of a VM address and the address must be valid. The
+ * "out_addr" parameter contains the address if successful.
+ *
+ * Results: 0 on success, EFAULT if the length is too small. See
+ * vsock_addr_validate() for other possible return codes.
+ *
+ * Side effects: None.
+ */
+
+int vsock_addr_cast(const struct sockaddr *addr,
+ size_t len, struct sockaddr_vm **out_addr)
+{
+ BUG_ON(!out_addr);
+
+ if (len < sizeof **out_addr)
+ return -EFAULT;
+
+ *out_addr = (struct sockaddr_vm *)addr;
+ return vsock_addr_validate(*out_addr);
+}
+
+/*
+ *
+ * vsock_addr_socket_context_stream --
+ *
+ * Determines whether the provided context id represents a context that
+ * contains a stream socket endpoints.
+ *
+ * Results: TRUE if the context does have socket endpoints, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_addr_socket_context_stream(u32 cid)
+{
+ static const vmci_id non_socket_contexts[] = {
+ VMCI_HYPERVISOR_CONTEXT_ID,
+ VMCI_WELL_KNOWN_CONTEXT_ID,
+ };
+ int i;
+
+ BUILD_BUG_ON(sizeof cid != sizeof *non_socket_contexts);
+
+ for (i = 0; i < ARRAY_SIZE(non_socket_contexts); i++) {
+ if (cid == non_socket_contexts[i])
+ return false;
+
+ }
+
+ return true;
+}
+
+/*
+ *
+ * vsock_addr_socket_context_dgram --
+ *
+ * Determines whether the provided <context id, resource id> represent a
+ * protected datagram endpoint.
+ *
+ * Results: TRUE if the context does have socket endpoints, FALSE otherwise.
+ *
+ * Side effects: None.
+ */
+
+bool vsock_addr_socket_context_dgram(u32 cid, u32 rid)
+{
+ if (cid == VMCI_HYPERVISOR_CONTEXT_ID) {
+ /*
+ * Registrations of PBRPC Servers do not modify VMX/Hypervisor
+ * state and are allowed.
+ */
+ return rid == VMCI_UNITY_PBRPC_REGISTER;
+ }
+
+ return true;
+}
diff --git a/net/vmw_vsock/vsock_addr.h b/net/vmw_vsock/vsock_addr.h
new file mode 100644
index 0000000..18f023d
--- /dev/null
+++ b/net/vmw_vsock/vsock_addr.h
@@ -0,0 +1,40 @@
+/*
+ * VMware vSockets Driver
+ *
+ * Copyright (C) 2007-2012 VMware, Inc. All rights reserved.
+ *
+ * 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 version 2 and no later version.
+ *
+ * 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.
+ */
+
+/*
+ * vsockAddr.h --
+ *
+ * VSockets address constants, types and functions.
+ */
+
+#ifndef _VSOCK_ADDR_H_
+#define _VSOCK_ADDR_H_
+
+void vsock_addr_init(struct sockaddr_vm *addr, u32 cid, u32 port);
+int vsock_addr_validate(const struct sockaddr_vm *addr);
+bool vsock_addr_bound(const struct sockaddr_vm *addr);
+void vsock_addr_unbind(struct sockaddr_vm *addr);
+bool vsock_addr_equals_addr(const struct sockaddr_vm *addr,
+ const struct sockaddr_vm *other);
+bool vsock_addr_equals_addr_any(const struct sockaddr_vm *addr,
+ const struct sockaddr_vm *other);
+bool vsock_addr_equals_handle_port(const struct sockaddr_vm *addr,
+ struct vmci_handle handle, u32 port);
+int vsock_addr_cast(const struct sockaddr *addr, size_t len,
+ struct sockaddr_vm **out_addr);
+bool vsock_addr_socket_context_stream(u32 cid);
+bool vsock_addr_socket_context_dgram(u32 cid, u32 rid);
+
+#endif
The big and only question is whether anyone can actually use any of
this stuff without your proprietary bits?
From: David Miller <[email protected]>
Date: Mon, 05 Nov 2012 13:09:17 -0500 (EST)
> The big and only question is whether anyone can actually use any of
> this stuff without your proprietary bits?
And BTW [email protected] bounces, take it out of the CC: list
on all future emails.
Hi David,
> The big and only question is whether anyone can actually use any of
> this stuff without your proprietary bits?
Do you mean the VMCI calls? The VMCI driver is in the process of being
upstreamed into the drivers/misc tree. Greg (cc'd on these patches) is
actively reviewing that code and we are addressing feedback.
Also, there was some interest from RedHat into using vSockets as a unified
interface, routed over a hypervisor-specific transport (virtio or
otherwise, although for now VMCI is the only one implemented).
Thanks!
- Andy
Hi George,
On Mon, Nov 5, 2012 at 1:00 PM, George Zhang <[email protected]> wrote:
> For additional information about the use of VMCI and in particular
> VMCI Sockets, please refer to the VMCI Socket Programming Guide
> available at https://www.vmware.com/support/developer/vmci-sdk/.
Is there documentation about the protocol itself? I couldn't find it
on the link above.
Thanks,
Sasha
David,
We will fix this bounces and remove the email: [email protected].
Thanks,
george zhang
----- Original Message -----
From: "David Miller" <[email protected]>
To: [email protected]
Cc: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
Sent: Monday, November 5, 2012 10:10:37 AM
Subject: Re: [PATCH 0/6] VSOCK for Linux upstreaming
From: David Miller <[email protected]>
Date: Mon, 05 Nov 2012 13:09:17 -0500 (EST)
> The big and only question is whether anyone can actually use any of
> this stuff without your proprietary bits?
And BTW [email protected] bounces, take it out of the CC: list
on all future emails.
Sasha,
There is a 'brief' implementation notes aboutVSOCK protocol family
in af_vsock.c. More about user space layer socket programming is from
VMCI Socket Programming Guide. (remove [email protected] for email bouncing).
thanks,
George
----- Original Message -----
From: "Sasha Levin" <[email protected]>
To: "George Zhang" <[email protected]>
Cc: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
Sent: Monday, November 5, 2012 10:23:02 AM
Subject: Re: [PATCH 0/6] VSOCK for Linux upstreaming
Hi George,
On Mon, Nov 5, 2012 at 1:00 PM, George Zhang <[email protected]> wrote:
> For additional information about the use of VMCI and in particular
> VMCI Sockets, please refer to the VMCI Socket Programming Guide
> available at https://www.vmware.com/support/developer/vmci-sdk/.
Is there documentation about the protocol itself? I couldn't find it
on the link above.
Thanks,
Sasha
On Mon, 05 Nov 2012 10:00:38 -0800
George Zhang <[email protected]> wrote:
> * * *
> This series of VSOCK linux upstreaming patches include latest udpate from
> VMware.
>
> Summary of changes:
> - Add include/linux/socket.h for AF_VSOCK.
> - Cleanup some comments.
> - Cleanup makefiles.
>
>
>
> * * *
>
> In an effort to improve the out-of-the-box experience with Linux
> kernels for VMware users, VMware is working on readying the Virtual
> Machine Communication Interface (vmw_vmci) and VMCI Sockets (VSOCK)
> (vmw_vsock) kernel modules for inclusion in the Linux kernel. The
> purpose of this post is to acquire feedback on the vmw_vsock kernel
> module. The vmw_vmci kernel module has been presented in an early post.
>
>
> * * *
>
> VMCI Sockets allows virtual machines to communicate with host kernel
> modules and the VMware hypervisors. VMCI Sockets kernel module has
> dependency on VMCI kernel module. User level applications both in
> a virtual machine and on the host can use vmw_vmci through VMCI
> Sockets API which facilitates fast and efficient communication
> between guest virtual machines and their host. A socket
> address family designed to be compatible with UDP and TCP at the
> interface level. Today, VMCI and VMCI Sockets are used by the VMware
> shared folders (HGFS) and various VMware Tools components inside the
> guest for zero-config, network-less access to VMware host services. In
> addition to this, VMware's users are using VMCI Sockets for various
> applications, where network access of the virtual machine is
> restricted or non-existent. Examples of this are VMs communicating
> with device proxies for proprietary hardware running as host
> applications and automated testing of applications running within
> virtual machines.
>
> The VMware VMCI Sockets are similar to other socket types, like
> Berkeley UNIX socket interface. The VMCI sockets module supports
> both connection-oriented stream sockets like TCP, and connectionless
> datagram sockets like UDP. The VSOCK protocol family is defined as
> "AF_VSOCK" and the socket operations split for SOCK_DGRAM and
> SOCK_STREAM.
>
> For additional information about the use of VMCI and in particular
> VMCI Sockets, please refer to the VMCI Socket Programming Guide
> available at https://www.vmware.com/support/developer/vmci-sdk/.
>
This should go to netdev as well since it is a new address family.
Never mind, mail server seemed to be overloaded today.
On Mon, 05 Nov 2012 10:00:51 -0800
George Zhang <[email protected]> wrote:
> + /* Added in 2.6.10. */
> + .owner = THIS_MODULE,
Thanks for submitting this, it will make life easier for distro's
that now have to go through extra effort to include out of mainline
support for Vmware.
You did some scrubbing of the macro's to support multiple kernel
versions, but there are still some leftovers.
This code seems to have a lot of "this added in version xxx"
type comments. These are probably not a good idea to include
in the mainline kernel code.
Hi Stephen,
> You did some scrubbing of the macro's to support multiple kernel
> versions, but there are still some leftovers.
> This code seems to have a lot of "this added in version xxx"
> type comments. These are probably not a good idea to include
> in the mainline kernel code.
Thanks so much for taking a look. Sorry about that, we'll remove
all such occurrences and send out a new series of patches in a bit.
Thanks!
- Andy
On 11/05/12 19:19, Andy King wrote:
> Hi David,
>
>> The big and only question is whether anyone can actually use any of
>> this stuff without your proprietary bits?
>
> Do you mean the VMCI calls? The VMCI driver is in the process of being
> upstreamed into the drivers/misc tree. Greg (cc'd on these patches) is
> actively reviewing that code and we are addressing feedback.
>
> Also, there was some interest from RedHat into using vSockets as a unified
> interface, routed over a hypervisor-specific transport (virtio or
> otherwise, although for now VMCI is the only one implemented).
Can you outline how this can be done? From a quick look over the code
it seems like vsock has a hard dependency on vmci, is that correct?
When making vsock a generic, reusable kernel service it should be the
other way around: vsock should provide the core implementation and an
interface where hypervisor-specific transports (vmci, virtio, xenbus,
...) can register themself.
cheers,
Gerd
Hi Gerd,
>> Also, there was some interest from RedHat into using vSockets as
>> a unified interface, routed over a hypervisor-specific transport
>> (virtio or otherwise, although for now VMCI is the only one
>> implemented).
>
> Can you outline how this can be done? From a quick look over the
> code it seems like vsock has a hard dependency on vmci, is that
> correct?
That's correct, VMCI is wired into vSockets and we don't currently
provide any way to insert another transport.
> When making vsock a generic, reusable kernel service it should be
> the other way around: vsock should provide the core implementation
> and an interface where hypervisor-specific transports (vmci,
> virtio, xenbus, ...) can register themself.
Sorry, that was a bad explanation on my part. You're absolutely
correct as to how it _should_ work. But it's up to Red Hat or others
to get the ball rolling and motivate the necessary work on vSockets
to make this happen. As Greg says, "everyone is lazy and just wants
their code accepted" ;)
Thanks!
- Andy
On 11/05/2012 01:32 PM, George Zhang wrote:
> Sasha,
> There is a 'brief' implementation notes aboutVSOCK protocol family
> in af_vsock.c. More about user space layer socket programming is from
> VMCI Socket Programming Guide. (remove [email protected] for email bouncing).
> thanks,
> George
So all the documentation I see in the VMCI Socket Programming Guide is about
userspace programming, and the documentation in af_vsock.c is all around
implementation considerations.
Is there anything that documents the actual protocol? Just for example, whats
the deal with REQUEST/REQUEST2? it appears like something to deal with legacy
code, but I'd really like to have it documented somewhere instead of trying
to figure how everything works by looking at the code.
Thanks,
Sasha
Hi Sasha,
Thanks for taking a look.
> So all the documentation I see in the VMCI Socket Programming Guide
> is about userspace programming, and the documentation in af_vsock.c
> is all around implementation considerations.
Agreed, we're sorely lacking in proper documentation for the internal
protocol. We're in the process of writing such a specification and
will post it to LKML next week at the latest.
> example, whats the deal with REQUEST/REQUEST2? it appears like
> something to deal with legacy code, but I'd really like to have it
> documented somewhere instead of trying to figure how everything
Correct, we have a legacy protocol and a v2, the latter now being
the default. This particular packet is sent by the client when
initiating a STREAM connection. The sequence is REQUEST(2)->
NEGOTIATE(2)->OFFER->ACCEPT. It will be properly documented in
the specification we post next week.
Thanks!
- Andy
On 11/07/2012 12:58 AM, Gerd Hoffmann wrote:
> On 11/05/12 19:19, Andy King wrote:
>> Hi David,
>>
>>> The big and only question is whether anyone can actually use any of
>>> this stuff without your proprietary bits?
>>
>> Do you mean the VMCI calls? The VMCI driver is in the process of being
>> upstreamed into the drivers/misc tree. Greg (cc'd on these patches) is
>> actively reviewing that code and we are addressing feedback.
>>
>> Also, there was some interest from RedHat into using vSockets as a unified
>> interface, routed over a hypervisor-specific transport (virtio or
>> otherwise, although for now VMCI is the only one implemented).
>
> Can you outline how this can be done? From a quick look over the code
> it seems like vsock has a hard dependency on vmci, is that correct?
>
> When making vsock a generic, reusable kernel service it should be the
> other way around: vsock should provide the core implementation and an
> interface where hypervisor-specific transports (vmci, virtio, xenbus,
> ...) can register themself.
This was already done in a hypervisor neutral way using virtio:
http://lists.openwall.net/netdev/2008/12/14/8
The concept was Nacked and that led to the abomination of virtio-serial. If an
address family for virtualization is on the table, we should reconsider
AF_VMCHANNEL.
I'd be thrilled to get rid of virtio-serial...
Regards,
Anthony Liguori
>
> cheers,
> Gerd
On Thu, 2012-11-15 at 15:32 -0600, Anthony Liguori wrote:
>
> The concept was Nacked and that led to the abomination of virtio-serial. If an
> address family for virtualization is on the table, we should reconsider
> AF_VMCHANNEL.
>
> I'd be thrilled to get rid of virtio-serial...
Ack.
Ben.