Add support for VSOCK to engine/net.c

* configure: add option to enable/disable vsock support

* engines/net.c: add vsock support

The  VSOCK  address family facilitates communication between virtual
machines and the host they are running on.

The addressing is formed by 2 integers: <CID, port>
- CID: Context ID, it is the ID assigned to the VM
  0, 1, 2 CIDs are reserved:
  0 - hypervisor CID (rarely used)
  1 - local communication (loopback)
  2 - host CID (the guest can always reach the host using CID=2)

- port: port number on 32bit to reach a specific process

* examples: add 3 simple job files for vsock (one sender, one receiver
  and one that uses vsock loopback interface similar to
  examples/netio.fio)

* fio.1: add vsock to supported protocols together with the required
  parameters

* HOWTO.rst: add vsock to supported protocols together with the required
  parameters

Signed-off-by: Marco Pinna <marco.pinn95@gmail.com>

diff --git a/HOWTO.rst b/HOWTO.rst
index ba160551c0f83e1d58f7c5c049e83e2472287e2d..53b03021d4aadb94e2da2ed30837a131ab696b3b 100644 (file)
--- a/HOWTO.rst
+++ b/HOWTO.rst
@@ -2626,10 +2626,13 @@ with the caveat that when used on the command line, they must come after the
                User datagram protocol V6.
        **unix**
                UNIX domain socket.
+       **vsock**
+               VSOCK protocol.
 
-       When the protocol is TCP or UDP, the port must also be given, as well as the
-       hostname if the job is a TCP listener or UDP reader. For unix sockets, the
+       When the protocol is TCP, UDP or VSOCK, the port must also be given, as well as the
+       hostname if the job is a TCP or VSOCK listener or UDP reader. For unix sockets, the
        normal :option:`filename` option should be used and the port is invalid.
+       When the protocol is VSOCK, the :option:`hostname` is the CID of the remote VM.
 
 .. option:: listen : [netsplice] [net]
 

diff --git a/configure b/configure
index dea8d07d113bbee100794dc1111bf4820a7b81b8..becb193ec8da836ce9131c95da35208aa9b2d286 100755 (executable)
--- a/configure
+++ b/configure
@@ -1728,6 +1728,25 @@ elif compile_prog "" "-lws2_32" "TCP_NODELAY"; then
 fi
 print_config "TCP_NODELAY" "$tcp_nodelay"
 
+##########################################
+# Check whether we have vsock
+if test "$vsock" != "yes" ; then
+  vsock="no"
+fi
+cat > $TMPC << EOF
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <linux/vm_sockets.h>
+int main(int argc, char **argv)
+{
+  return socket(AF_VSOCK, SOCK_STREAM, 0);
+}
+EOF
+if compile_prog "" "" "vsock"; then
+  vsock="yes"
+fi
+print_config "vsock" "$vsock"
+
 ##########################################
 # Check whether we have SO_SNDBUF
 if test "$window_size" != "yes" ; then
@@ -3192,6 +3211,9 @@ fi
 if test "$ipv6" = "yes" ; then
   output_sym "CONFIG_IPV6"
 fi
+if test "$vsock" = "yes"; then
+  output_sym "CONFIG_VSOCK"
+fi
 if test "$http" = "yes" ; then
   output_sym "CONFIG_HTTP"
 fi

diff --git a/engines/net.c b/engines/net.c
index fec53d74175eedb5d4dd3be6e7a68ece50bb629f..29150bb348ac8b2342f2441c1198ce21e5499fa1 100644 (file)
--- a/engines/net.c
+++ b/engines/net.c
@@ -18,6 +18,16 @@
 #include <sys/socket.h>
 #include <sys/un.h>
 
+#ifdef CONFIG_VSOCK
+#include <linux/vm_sockets.h>
+#else
+struct sockaddr_vm {
+};
+#ifndef AF_VSOCK
+#define AF_VSOCK       -1
+#endif
+#endif
+
 #include "../fio.h"
 #include "../verify.h"
 #include "../optgroup.h"
@@ -30,6 +40,7 @@ struct netio_data {
        struct sockaddr_in addr;
        struct sockaddr_in6 addr6;
        struct sockaddr_un addr_un;
+       struct sockaddr_vm addr_vm;
        uint64_t udp_send_seq;
        uint64_t udp_recv_seq;
 };
@@ -69,6 +80,7 @@ enum {
        FIO_TYPE_UNIX   = 3,
        FIO_TYPE_TCP_V6 = 4,
        FIO_TYPE_UDP_V6 = 5,
+       FIO_TYPE_VSOCK_STREAM   = 6,
 };
 
 static int str_hostname_cb(void *data, const char *input);
@@ -126,6 +138,10 @@ static struct fio_option options[] = {
                            .oval = FIO_TYPE_UNIX,
                            .help = "UNIX domain socket",
                          },
+                         { .ival = "vsock",
+                           .oval = FIO_TYPE_VSOCK_STREAM,
+                           .help = "Virtual socket",
+                         },
                },
                .category = FIO_OPT_C_ENGINE,
                .group  = FIO_OPT_G_NETIO,
@@ -223,6 +239,11 @@ static inline int is_ipv6(struct netio_options *o)
        return o->proto == FIO_TYPE_UDP_V6 || o->proto == FIO_TYPE_TCP_V6;
 }
 
+static inline int is_vsock(struct netio_options *o)
+{
+       return o->proto == FIO_TYPE_VSOCK_STREAM;
+}
+
 static int set_window_size(struct thread_data *td, int fd)
 {
 #ifdef CONFIG_NET_WINDOWSIZE
@@ -732,6 +753,9 @@ static int fio_netio_connect(struct thread_data *td, struct fio_file *f)
        } else if (o->proto == FIO_TYPE_UNIX) {
                domain = AF_UNIX;
                type = SOCK_STREAM;
+       } else if (is_vsock(o)) {
+               domain = AF_VSOCK;
+               type = SOCK_STREAM;
        } else {
                log_err("fio: bad network type %d\n", o->proto);
                f->fd = -1;
@@ -809,7 +833,14 @@ static int fio_netio_connect(struct thread_data *td, struct fio_file *f)
                        close(f->fd);
                        return 1;
                }
+       } else if (is_vsock(o)) {
+               socklen_t len = sizeof(nd->addr_vm);
 
+               if (connect(f->fd, (struct sockaddr *) &nd->addr_vm, len) < 0) {
+                       td_verror(td, errno, "connect");
+                       close(f->fd);
+                       return 1;
+               }
        } else {
                struct sockaddr_un *addr = &nd->addr_un;
                socklen_t len;
@@ -849,6 +880,9 @@ static int fio_netio_accept(struct thread_data *td, struct fio_file *f)
        if (o->proto == FIO_TYPE_TCP) {
                socklen = sizeof(nd->addr);
                f->fd = accept(nd->listenfd, (struct sockaddr *) &nd->addr, &socklen);
+       } else if (is_vsock(o)) {
+               socklen = sizeof(nd->addr_vm);
+               f->fd = accept(nd->listenfd, (struct sockaddr *) &nd->addr_vm, &socklen);
        } else {
                socklen = sizeof(nd->addr6);
                f->fd = accept(nd->listenfd, (struct sockaddr *) &nd->addr6, &socklen);
@@ -890,6 +924,9 @@ static void fio_netio_send_close(struct thread_data *td, struct fio_file *f)
        if (is_ipv6(o)) {
                to = (struct sockaddr *) &nd->addr6;
                len = sizeof(nd->addr6);
+       } else if (is_vsock(o)) {
+               to = NULL;
+               len = 0;
        } else {
                to = (struct sockaddr *) &nd->addr;
                len = sizeof(nd->addr);
@@ -960,6 +997,9 @@ static int fio_netio_send_open(struct thread_data *td, struct fio_file *f)
        if (is_ipv6(o)) {
                len = sizeof(nd->addr6);
                to = (struct sockaddr *) &nd->addr6;
+       } else if (is_vsock(o)) {
+               len = sizeof(nd->addr_vm);
+               to = (struct sockaddr *) &nd->addr_vm;
        } else {
                len = sizeof(nd->addr);
                to = (struct sockaddr *) &nd->addr;
@@ -1023,13 +1063,17 @@ static int fio_fill_addr(struct thread_data *td, const char *host, int af,
 
        memset(&hints, 0, sizeof(hints));
 
-       if (is_tcp(o))
+       if (is_tcp(o) || is_vsock(o))
                hints.ai_socktype = SOCK_STREAM;
        else
                hints.ai_socktype = SOCK_DGRAM;
 
        if (is_ipv6(o))
                hints.ai_family = AF_INET6;
+#ifdef CONFIG_VSOCK
+       else if (is_vsock(o))
+               hints.ai_family = AF_VSOCK;
+#endif
        else
                hints.ai_family = AF_INET;
 
@@ -1110,12 +1154,50 @@ static int fio_netio_setup_connect_unix(struct thread_data *td,
        return 0;
 }
 
+static int fio_netio_setup_connect_vsock(struct thread_data *td,
+                                       const char *host, unsigned short port)
+{
+#ifdef CONFIG_VSOCK
+       struct netio_data *nd = td->io_ops_data;
+       struct sockaddr_vm *addr = &nd->addr_vm;
+       int cid;
+
+       if (!host) {
+               log_err("fio: connect with no host to connect to.\n");
+               if (td_read(td))
+                       log_err("fio: did you forget to set 'listen'?\n");
+
+               td_verror(td, EINVAL, "no hostname= set");
+               return 1;
+       }
+
+       addr->svm_family = AF_VSOCK;
+       addr->svm_port = port;
+
+       if (host) {
+               cid = atoi(host);
+               if (cid < 0 || cid > UINT32_MAX) {
+                       log_err("fio: invalid CID %d\n", cid);
+                       return 1;
+               }
+               addr->svm_cid = cid;
+       }
+
+       return 0;
+#else
+       td_verror(td, -EINVAL, "vsock not supported");
+       return 1;
+#endif
+}
+
 static int fio_netio_setup_connect(struct thread_data *td)
 {
        struct netio_options *o = td->eo;
 
        if (is_udp(o) || is_tcp(o))
                return fio_netio_setup_connect_inet(td, td->o.filename,o->port);
+       else if (is_vsock(o))
+               return fio_netio_setup_connect_vsock(td, td->o.filename, o->port);
        else
                return fio_netio_setup_connect_unix(td, td->o.filename);
 }
@@ -1268,6 +1350,47 @@ static int fio_netio_setup_listen_inet(struct thread_data *td, short port)
        return 0;
 }
 
+static int fio_netio_setup_listen_vsock(struct thread_data *td, short port, int type)
+{
+#ifdef CONFIG_VSOCK
+       struct netio_data *nd = td->io_ops_data;
+       struct sockaddr_vm *addr = &nd->addr_vm;
+       int fd, opt;
+       socklen_t len;
+
+       fd = socket(AF_VSOCK, type, 0);
+       if (fd < 0) {
+               td_verror(td, errno, "socket");
+               return 1;
+       }
+
+       opt = 1;
+       if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (void *) &opt, sizeof(opt)) < 0) {
+               td_verror(td, errno, "setsockopt");
+               close(fd);
+               return 1;
+       }
+
+       len = sizeof(*addr);
+
+       nd->addr_vm.svm_family = AF_VSOCK;
+       nd->addr_vm.svm_cid = VMADDR_CID_ANY;
+       nd->addr_vm.svm_port = port;
+
+       if (bind(fd, (struct sockaddr *) addr, len) < 0) {
+               td_verror(td, errno, "bind");
+               close(fd);
+               return 1;
+       }
+
+       nd->listenfd = fd;
+       return 0;
+#else
+       td_verror(td, -EINVAL, "vsock not supported");
+       return -1;
+#endif
+}
+
 static int fio_netio_setup_listen(struct thread_data *td)
 {
        struct netio_data *nd = td->io_ops_data;
@@ -1276,6 +1399,8 @@ static int fio_netio_setup_listen(struct thread_data *td)
 
        if (is_udp(o) || is_tcp(o))
                ret = fio_netio_setup_listen_inet(td, o->port);
+       else if (is_vsock(o))
+               ret = fio_netio_setup_listen_vsock(td, o->port, SOCK_STREAM);
        else
                ret = fio_netio_setup_listen_unix(td, td->o.filename);
 
@@ -1311,6 +1436,9 @@ static int fio_netio_init(struct thread_data *td)
        if (o->proto == FIO_TYPE_UNIX && o->port) {
                log_err("fio: network IO port not valid with unix socket\n");
                return 1;
+       } else if (is_vsock(o) && !o->port) {
+               log_err("fio: network IO requires port for vsock\n");
+               return 1;
        } else if (o->proto != FIO_TYPE_UNIX && !o->port) {
                log_err("fio: network IO requires port for tcp or udp\n");
                return 1;
@@ -1318,7 +1446,7 @@ static int fio_netio_init(struct thread_data *td)
 
        o->port += td->subjob_number;
 
-       if (!is_tcp(o)) {
+       if (!is_tcp(o) && !is_vsock(o)) {
                if (o->listen) {
                        log_err("fio: listen only valid for TCP proto IO\n");
                        return 1;

diff --git a/examples/netio_vsock.fio b/examples/netio_vsock.fio
new file mode 100644 (file)
index 0000000..8c328f7
--- /dev/null
+++ b/examples/netio_vsock.fio
@@ -0,0 +1,22 @@
+# Example network vsock job, just defines two clients that send/recv data
+[global]
+ioengine=net
+
+port=8888
+protocol=vsock
+bs=4k
+size=100g
+
+#set the below option to enable end-to-end data integrity tests
+#verify=md5
+
+[receiver]
+listen
+rw=read
+
+[sender]
+# 1 (VMADDR_CID_LOCAL) is the well-known address
+# for local communication (loopback)
+hostname=1
+startdelay=1
+rw=write

diff --git a/examples/netio_vsock_receiver.fio b/examples/netio_vsock_receiver.fio
new file mode 100644 (file)
index 0000000..e2a00c4
--- /dev/null
+++ b/examples/netio_vsock_receiver.fio
@@ -0,0 +1,14 @@
+# Example network vsock job, just defines a receiver
+[global]
+ioengine=net
+port=8888
+protocol=vsock
+bs=4k
+size=100g
+
+#set the below option to enable end-to-end data integrity tests
+#verify=md5
+
+[receiver]
+listen
+rw=read

diff --git a/examples/netio_vsock_sender.fio b/examples/netio_vsock_sender.fio
new file mode 100644 (file)
index 0000000..2451d99
--- /dev/null
+++ b/examples/netio_vsock_sender.fio
@@ -0,0 +1,17 @@
+# Example network vsock job, just defines a sender
+[global]
+ioengine=net
+port=8888
+protocol=vsock
+bs=4k
+size=100g
+
+#set the below option to enable end-to-end data integrity tests
+#verify=md5
+
+[sender]
+# set the 'hostname' option to the CID of the listening domain
+hostname=3
+startdelay=1
+rw=write
+

diff --git a/fio.1 b/fio.1
index aef1dc855e18f65343fae352e92b7908d49c599f..227fcb4786c991bb193d8645c7e3f898df5eb3f1 100644 (file)
--- a/fio.1
+++ b/fio.1
@@ -2376,11 +2376,16 @@ User datagram protocol V6.
 .TP
 .B unix
 UNIX domain socket.
+.TP
+.B vsock
+VSOCK protocol.
 .RE
 .P
-When the protocol is TCP or UDP, the port must also be given, as well as the
-hostname if the job is a TCP listener or UDP reader. For unix sockets, the
+When the protocol is TCP, UDP or VSOCK, the port must also be given, as well as the
+hostname if the job is a TCP or VSOCK listener or UDP reader. For unix sockets, the
 normal \fBfilename\fR option should be used and the port is invalid.
+When the protocol is VSOCK, the \fBhostname\fR is the CID of the remote VM.
+
 .RE
 .TP
 .BI (netsplice,net)listen