#include <arpa/inet.h>
#include <netdb.h>
#include <sys/poll.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <sys/un.h>
#include "../fio.h"
-#include "../os.h"
struct netio_data {
int listenfd;
int send_to_net;
+ int use_splice;
+ int type;
+ int pipes[2];
char host[64];
struct sockaddr_in addr;
+ struct sockaddr_un addr_un;
};
+struct udp_close_msg {
+ uint32_t magic;
+ uint32_t cmd;
+};
+
+enum {
+ FIO_LINK_CLOSE = 0x89,
+ FIO_LINK_CLOSE_MAGIC = 0x6c696e6b,
+
+ FIO_TYPE_TCP = 1,
+ FIO_TYPE_UDP = 2,
+ FIO_TYPE_UNIX = 3,
+};
+
+/*
+ * Return -1 for error and 'nr events' for a positive number
+ * of events
+ */
+static int poll_wait(struct thread_data *td, int fd, short events)
+{
+ struct pollfd pfd;
+ int ret;
+
+ while (!td->terminate) {
+ pfd.fd = fd;
+ pfd.events = events;
+ ret = poll(&pfd, 1, -1);
+ if (ret < 0) {
+ if (errno == EINTR)
+ break;
+
+ td_verror(td, errno, "poll");
+ return -1;
+ } else if (!ret)
+ continue;
+
+ break;
+ }
+
+ if (pfd.revents & events)
+ return 1;
+
+ return -1;
+}
+
static int fio_netio_prep(struct thread_data *td, struct io_u *io_u)
{
struct netio_data *nd = td->io_ops->data;
- struct fio_file *f = io_u->file;
/*
* Make sure we don't see spurious reads to a receiver, and vice versa
return 1;
}
- if (io_u->ddir == DDIR_SYNC)
+ return 0;
+}
+
+#ifdef FIO_HAVE_SPLICE
+static int splice_io_u(int fdin, int fdout, unsigned int len)
+{
+ int bytes = 0;
+
+ while (len) {
+ int ret = splice(fdin, NULL, fdout, NULL, len, 0);
+
+ if (ret < 0) {
+ if (!bytes)
+ bytes = ret;
+
+ break;
+ } else if (!ret)
+ break;
+
+ bytes += ret;
+ len -= ret;
+ }
+
+ return bytes;
+}
+
+/*
+ * Receive bytes from a socket and fill them into the internal pipe
+ */
+static int splice_in(struct thread_data *td, struct io_u *io_u)
+{
+ struct netio_data *nd = td->io_ops->data;
+
+ return splice_io_u(io_u->file->fd, nd->pipes[1], io_u->xfer_buflen);
+}
+
+/*
+ * Transmit 'len' bytes from the internal pipe
+ */
+static int splice_out(struct thread_data *td, struct io_u *io_u,
+ unsigned int len)
+{
+ struct netio_data *nd = td->io_ops->data;
+
+ return splice_io_u(nd->pipes[0], io_u->file->fd, len);
+}
+
+static int vmsplice_io_u(struct io_u *io_u, int fd, unsigned int len)
+{
+ struct iovec iov = {
+ .iov_base = io_u->xfer_buf,
+ .iov_len = len,
+ };
+ int bytes = 0;
+
+ while (iov.iov_len) {
+ int ret = vmsplice(fd, &iov, 1, SPLICE_F_MOVE);
+
+ if (ret < 0) {
+ if (!bytes)
+ bytes = ret;
+ break;
+ } else if (!ret)
+ break;
+
+ iov.iov_len -= ret;
+ iov.iov_base += ret;
+ bytes += ret;
+ }
+
+ return bytes;
+
+}
+
+/*
+ * vmsplice() pipe to io_u buffer
+ */
+static int vmsplice_io_u_out(struct thread_data *td, struct io_u *io_u,
+ unsigned int len)
+{
+ struct netio_data *nd = td->io_ops->data;
+
+ return vmsplice_io_u(io_u, nd->pipes[0], len);
+}
+
+/*
+ * vmsplice() io_u to pipe
+ */
+static int vmsplice_io_u_in(struct thread_data *td, struct io_u *io_u)
+{
+ struct netio_data *nd = td->io_ops->data;
+
+ return vmsplice_io_u(io_u, nd->pipes[1], io_u->xfer_buflen);
+}
+
+/*
+ * splice receive - transfer socket data into a pipe using splice, then map
+ * that pipe data into the io_u using vmsplice.
+ */
+static int fio_netio_splice_in(struct thread_data *td, struct io_u *io_u)
+{
+ int ret;
+
+ ret = splice_in(td, io_u);
+ if (ret > 0)
+ return vmsplice_io_u_out(td, io_u, ret);
+
+ return ret;
+}
+
+/*
+ * splice transmit - map data from the io_u into a pipe by using vmsplice,
+ * then transfer that pipe to a socket using splice.
+ */
+static int fio_netio_splice_out(struct thread_data *td, struct io_u *io_u)
+{
+ int ret;
+
+ ret = vmsplice_io_u_in(td, io_u);
+ if (ret > 0)
+ return splice_out(td, io_u, ret);
+
+ return ret;
+}
+#else
+static int fio_netio_splice_in(struct thread_data *td, struct io_u *io_u)
+{
+ errno = EOPNOTSUPP;
+ return -1;
+}
+
+static int fio_netio_splice_out(struct thread_data *td, struct io_u *io_u)
+{
+ errno = EOPNOTSUPP;
+ return -1;
+}
+#endif
+
+static int fio_netio_send(struct thread_data *td, struct io_u *io_u)
+{
+ struct netio_data *nd = td->io_ops->data;
+ int ret, flags = OS_MSG_DONTWAIT;
+
+ do {
+ if (nd->type == FIO_TYPE_UDP) {
+ struct sockaddr *to = (struct sockaddr *) &nd->addr;
+
+ ret = sendto(io_u->file->fd, io_u->xfer_buf,
+ io_u->xfer_buflen, flags, to,
+ sizeof(*to));
+ } else {
+ /*
+ * if we are going to write more, set MSG_MORE
+ */
+#ifdef MSG_MORE
+ if (td->this_io_bytes[DDIR_WRITE] + io_u->xfer_buflen <
+ td->o.size)
+ flags |= MSG_MORE;
+#endif
+ ret = send(io_u->file->fd, io_u->xfer_buf,
+ io_u->xfer_buflen, flags);
+ }
+ if (ret > 0)
+ break;
+
+ ret = poll_wait(td, io_u->file->fd, POLLOUT);
+ if (ret <= 0)
+ break;
+
+ flags &= ~OS_MSG_DONTWAIT;
+ } while (1);
+
+ return ret;
+}
+
+static int is_udp_close(struct io_u *io_u, int len)
+{
+ struct udp_close_msg *msg;
+
+ if (len != sizeof(struct udp_close_msg))
+ return 0;
+
+ msg = io_u->xfer_buf;
+ if (ntohl(msg->magic) != FIO_LINK_CLOSE_MAGIC)
return 0;
- if (io_u->offset == f->last_completed_pos)
+ if (ntohl(msg->cmd) != FIO_LINK_CLOSE)
return 0;
- /*
- * If offset is different from last end position, it's a seek.
- * As network io is purely sequential, we don't allow seeks.
- */
- td_verror(td, EINVAL, "cannot seek");
return 1;
}
+static int fio_netio_recv(struct thread_data *td, struct io_u *io_u)
+{
+ struct netio_data *nd = td->io_ops->data;
+ int ret, flags = OS_MSG_DONTWAIT;
+
+ do {
+ if (nd->type == FIO_TYPE_UDP) {
+ fio_socklen_t len = sizeof(nd->addr);
+ struct sockaddr *from = (struct sockaddr *) &nd->addr;
+
+ ret = recvfrom(io_u->file->fd, io_u->xfer_buf,
+ io_u->xfer_buflen, flags, from, &len);
+ if (is_udp_close(io_u, ret)) {
+ td->done = 1;
+ return 0;
+ }
+ } else {
+ ret = recv(io_u->file->fd, io_u->xfer_buf,
+ io_u->xfer_buflen, flags);
+ }
+ if (ret > 0)
+ break;
+
+ ret = poll_wait(td, io_u->file->fd, POLLIN);
+ if (ret <= 0)
+ break;
+ flags &= ~OS_MSG_DONTWAIT;
+ flags |= MSG_WAITALL;
+ } while (1);
+
+ return ret;
+}
+
static int fio_netio_queue(struct thread_data *td, struct io_u *io_u)
{
- struct fio_file *f = io_u->file;
- int ret, flags = 0;
+ struct netio_data *nd = td->io_ops->data;
+ int ret;
+
+ fio_ro_check(td, io_u);
if (io_u->ddir == DDIR_WRITE) {
- /*
- * if we are going to write more, set MSG_MORE
- */
- if (td->this_io_bytes[DDIR_WRITE] + io_u->xfer_buflen <
- td->io_size)
- flags = MSG_MORE;
-
- ret = send(f->fd, io_u->xfer_buf, io_u->xfer_buflen, flags);
+ if (!nd->use_splice || nd->type == FIO_TYPE_UDP ||
+ nd->type == FIO_TYPE_UNIX)
+ ret = fio_netio_send(td, io_u);
+ else
+ ret = fio_netio_splice_out(td, io_u);
} else if (io_u->ddir == DDIR_READ) {
- flags = MSG_WAITALL;
- ret = recv(f->fd, io_u->xfer_buf, io_u->xfer_buflen, flags);
+ if (!nd->use_splice || nd->type == FIO_TYPE_UDP ||
+ nd->type == FIO_TYPE_UDP)
+ ret = fio_netio_recv(td, io_u);
+ else
+ ret = fio_netio_splice_in(td, io_u);
} else
ret = 0; /* must be a SYNC */
io_u->resid = io_u->xfer_buflen - ret;
io_u->error = 0;
return FIO_Q_COMPLETED;
- } else
- io_u->error = errno;
+ } else {
+ int err = errno;
+
+ if (io_u->ddir == DDIR_WRITE && err == EMSGSIZE)
+ return FIO_Q_BUSY;
+
+ io_u->error = err;
+ }
}
if (io_u->error)
static int fio_netio_connect(struct thread_data *td, struct fio_file *f)
{
struct netio_data *nd = td->io_ops->data;
+ int type, domain;
+
+ if (nd->type == FIO_TYPE_TCP) {
+ domain = AF_INET;
+ type = SOCK_STREAM;
+ } else if (nd->type == FIO_TYPE_UDP) {
+ domain = AF_INET;
+ type = SOCK_DGRAM;
+ } else if (nd->type == FIO_TYPE_UNIX) {
+ domain = AF_UNIX;
+ type = SOCK_STREAM;
+ } else {
+ log_err("fio: bad network type %d\n", nd->type);
+ f->fd = -1;
+ return 1;
+ }
- f->fd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+ f->fd = socket(domain, type, 0);
if (f->fd < 0) {
td_verror(td, errno, "socket");
return 1;
}
- if (connect(f->fd, (struct sockaddr *) &nd->addr, sizeof(nd->addr)) < 0) {
- td_verror(td, errno, "connect");
- return 1;
+ if (nd->type == FIO_TYPE_UDP)
+ return 0;
+ else if (nd->type == FIO_TYPE_TCP) {
+ fio_socklen_t len = sizeof(nd->addr);
+
+ if (connect(f->fd, (struct sockaddr *) &nd->addr, len) < 0) {
+ td_verror(td, errno, "connect");
+ return 1;
+ }
+ } else {
+ struct sockaddr_un *addr = &nd->addr_un;
+ fio_socklen_t len;
+
+ len = sizeof(addr->sun_family) + strlen(addr->sun_path) + 1;
+
+ if (connect(f->fd, (struct sockaddr *) addr, len) < 0) {
+ td_verror(td, errno, "connect");
+ return 1;
+ }
}
return 0;
static int fio_netio_accept(struct thread_data *td, struct fio_file *f)
{
struct netio_data *nd = td->io_ops->data;
- socklen_t socklen = sizeof(nd->addr);
- struct pollfd pfd;
- int ret;
-
- log_info("fio: waiting for connection\n");
+ fio_socklen_t socklen = sizeof(nd->addr);
- /*
- * Accept loop. poll for incoming events, accept them. Repeat until we
- * have all connections.
- */
- while (!td->terminate) {
- pfd.fd = nd->listenfd;
- pfd.events = POLLIN;
-
- ret = poll(&pfd, 1, -1);
- if (ret < 0) {
- if (errno == EINTR)
- continue;
+ if (nd->type == FIO_TYPE_UDP) {
+ f->fd = nd->listenfd;
+ return 0;
+ }
- td_verror(td, errno, "poll");
- break;
- } else if (!ret)
- continue;
+ log_info("fio: waiting for connection\n");
- /*
- * should be impossible
- */
- if (!(pfd.revents & POLLIN))
- continue;
+ if (poll_wait(td, nd->listenfd, POLLIN) < 0)
+ return 1;
- f->fd = accept(nd->listenfd, (struct sockaddr *) &nd->addr, &socklen);
- if (f->fd < 0) {
- td_verror(td, errno, "accept");
- return 1;
- }
- break;
+ f->fd = accept(nd->listenfd, (struct sockaddr *) &nd->addr, &socklen);
+ if (f->fd < 0) {
+ td_verror(td, errno, "accept");
+ return 1;
}
return 0;
}
-
static int fio_netio_open_file(struct thread_data *td, struct fio_file *f)
{
+ int ret;
+
if (td_read(td))
- return fio_netio_accept(td, f);
+ ret = fio_netio_accept(td, f);
else
- return fio_netio_connect(td, f);
+ ret = fio_netio_connect(td, f);
+
+ if (ret)
+ f->fd = -1;
+ return ret;
}
-static int fio_netio_setup_connect(struct thread_data *td, const char *host,
- unsigned short port)
+static void fio_netio_udp_close(struct thread_data *td, struct fio_file *f)
+{
+ struct netio_data *nd = td->io_ops->data;
+ struct udp_close_msg msg;
+ struct sockaddr *to = (struct sockaddr *) &nd->addr;
+ int ret;
+
+ msg.magic = htonl(FIO_LINK_CLOSE_MAGIC);
+ msg.cmd = htonl(FIO_LINK_CLOSE);
+
+ ret = sendto(f->fd, &msg, sizeof(msg), MSG_WAITALL, to,
+ sizeof(nd->addr));
+ if (ret < 0)
+ td_verror(td, errno, "sendto udp link close");
+}
+
+static int fio_netio_close_file(struct thread_data *td, struct fio_file *f)
+{
+ struct netio_data *nd = td->io_ops->data;
+
+ /*
+ * If this is an UDP connection, notify the receiver that we are
+ * closing down the link
+ */
+ if (nd->type == FIO_TYPE_UDP)
+ fio_netio_udp_close(td, f);
+
+ return generic_close_file(td, f);
+}
+
+static int fio_netio_setup_connect_inet(struct thread_data *td,
+ const char *host, unsigned short port)
{
struct netio_data *nd = td->io_ops->data;
return 0;
}
-static int fio_netio_setup_listen(struct thread_data *td, short port)
+static int fio_netio_setup_connect_unix(struct thread_data *td,
+ const char *path)
{
struct netio_data *nd = td->io_ops->data;
- int fd, opt;
+ struct sockaddr_un *soun = &nd->addr_un;
- fd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+ soun->sun_family = AF_UNIX;
+ strcpy(soun->sun_path, path);
+ return 0;
+}
+
+static int fio_netio_setup_connect(struct thread_data *td, const char *host,
+ unsigned short port)
+{
+ struct netio_data *nd = td->io_ops->data;
+
+ if (nd->type == FIO_TYPE_UDP || nd->type == FIO_TYPE_TCP)
+ return fio_netio_setup_connect_inet(td, host, port);
+ else
+ return fio_netio_setup_connect_unix(td, host);
+}
+
+static int fio_netio_setup_listen_unix(struct thread_data *td, const char *path)
+{
+ struct netio_data *nd = td->io_ops->data;
+ struct sockaddr_un *addr = &nd->addr_un;
+ mode_t mode;
+ int len, fd;
+
+ fd = socket(AF_UNIX, SOCK_STREAM, 0);
+ if (fd < 0) {
+ log_err("fio: socket: %s\n", strerror(errno));
+ return -1;
+ }
+
+ mode = umask(000);
+
+ memset(addr, 0, sizeof(*addr));
+ addr->sun_family = AF_UNIX;
+ strcpy(addr->sun_path, path);
+ unlink(path);
+
+ len = sizeof(addr->sun_family) + strlen(path) + 1;
+
+ if (bind(fd, (struct sockaddr *) addr, len) < 0) {
+ log_err("fio: bind: %s\n", strerror(errno));
+ return -1;
+ }
+
+ umask(mode);
+ nd->listenfd = fd;
+ return 0;
+}
+
+static int fio_netio_setup_listen_inet(struct thread_data *td, short port)
+{
+ struct netio_data *nd = td->io_ops->data;
+ int fd, opt, type;
+
+ if (nd->type == FIO_TYPE_TCP)
+ type = SOCK_STREAM;
+ else
+ type = SOCK_DGRAM;
+
+ fd = socket(AF_INET, type, 0);
if (fd < 0) {
td_verror(td, errno, "socket");
return 1;
td_verror(td, errno, "bind");
return 1;
}
- if (listen(fd, 1) < 0) {
+
+ nd->listenfd = fd;
+ return 0;
+}
+
+static int fio_netio_setup_listen(struct thread_data *td, const char *path,
+ short port)
+{
+ struct netio_data *nd = td->io_ops->data;
+ int ret;
+
+ if (nd->type == FIO_TYPE_UDP || nd->type == FIO_TYPE_TCP)
+ ret = fio_netio_setup_listen_inet(td, port);
+ else
+ ret = fio_netio_setup_listen_unix(td, path);
+
+ if (ret)
+ return ret;
+ if (nd->type == FIO_TYPE_UDP)
+ return 0;
+
+ if (listen(nd->listenfd, 10) < 0) {
td_verror(td, errno, "listen");
+ nd->listenfd = -1;
return 1;
}
- nd->listenfd = fd;
return 0;
}
static int fio_netio_init(struct thread_data *td)
{
struct netio_data *nd = td->io_ops->data;
- unsigned short port;
- struct fio_file *f;
+ unsigned int port;
char host[64], buf[128];
- unsigned int i;
- char *sep;
+ char *sep, *portp, *modep;
int ret;
- if (!td->total_file_size) {
- log_err("fio: need size= set\n");
- return 1;
- }
-
if (td_rw(td)) {
log_err("fio: network connections must be read OR write\n");
return 1;
}
-
- strcpy(buf, td->filename);
-
- sep = strchr(buf, '/');
- if (!sep) {
- log_err("fio: bad network host/port <<%s>>\n", td->filename);
+ if (td_random(td)) {
+ log_err("fio: network IO can't be random\n");
return 1;
}
+ strcpy(buf, td->o.filename);
+
+ sep = strchr(buf, ',');
+ if (!sep)
+ goto bad_host;
+
*sep = '\0';
sep++;
strcpy(host, buf);
- port = atoi(sep);
+ if (!strlen(host))
+ goto bad_host;
+
+ modep = NULL;
+ portp = sep;
+ sep = strchr(portp, ',');
+ if (sep) {
+ *sep = '\0';
+ modep = sep + 1;
+ }
+
+ if (!strncmp("tcp", modep, strlen(modep)) ||
+ !strncmp("TCP", modep, strlen(modep)))
+ nd->type = FIO_TYPE_TCP;
+ else if (!strncmp("udp", modep, strlen(modep)) ||
+ !strncmp("UDP", modep, strlen(modep)))
+ nd->type = FIO_TYPE_UDP;
+ else if (!strncmp("unix", modep, strlen(modep)) ||
+ !strncmp("UNIX", modep, strlen(modep)))
+ nd->type = FIO_TYPE_UNIX;
+ else
+ goto bad_host;
+
+ if (nd->type != FIO_TYPE_UNIX) {
+ port = strtol(portp, NULL, 10);
+ if (!port || port > 65535)
+ goto bad_host;
+ } else
+ port = 0;
if (td_read(td)) {
nd->send_to_net = 0;
- ret = fio_netio_setup_listen(td, port);
+ ret = fio_netio_setup_listen(td, host, port);
} else {
nd->send_to_net = 1;
ret = fio_netio_setup_connect(td, host, port);
}
- if (ret)
- return ret;
-
- td->io_size = td->total_file_size;
- td->total_io_size = td->io_size;
-
- for_each_file(td, f, i) {
- f->file_size = td->total_file_size / td->nr_files;
- f->real_file_size = f->file_size;
- }
-
- return 0;
+ return ret;
+bad_host:
+ log_err("fio: bad network host/port/protocol: %s\n", td->o.filename);
+ return 1;
}
static void fio_netio_cleanup(struct thread_data *td)
struct netio_data *nd = td->io_ops->data;
if (nd) {
+ if (nd->listenfd != -1)
+ close(nd->listenfd);
+ if (nd->pipes[0] != -1)
+ close(nd->pipes[0]);
+ if (nd->pipes[1] != -1)
+ close(nd->pipes[1]);
+
free(nd);
- td->io_ops->data = NULL;
}
}
static int fio_netio_setup(struct thread_data *td)
{
- struct netio_data *nd = malloc(sizeof(*nd));
+ struct netio_data *nd;
+
+ if (!td->io_ops->data) {
+ nd = malloc(sizeof(*nd));;
+
+ memset(nd, 0, sizeof(*nd));
+ nd->listenfd = -1;
+ nd->pipes[0] = nd->pipes[1] = -1;
+ td->io_ops->data = nd;
+ }
- memset(nd, 0, sizeof(*nd));
- nd->listenfd = -1;
- td->io_ops->data = nd;
return 0;
}
-static struct ioengine_ops ioengine = {
+#ifdef FIO_HAVE_SPLICE
+static int fio_netio_setup_splice(struct thread_data *td)
+{
+ struct netio_data *nd;
+
+ fio_netio_setup(td);
+
+ nd = td->io_ops->data;
+ if (nd) {
+ if (pipe(nd->pipes) < 0)
+ return 1;
+
+ nd->use_splice = 1;
+ return 0;
+ }
+
+ return 1;
+}
+
+static struct ioengine_ops ioengine_splice = {
+ .name = "netsplice",
+ .version = FIO_IOOPS_VERSION,
+ .prep = fio_netio_prep,
+ .queue = fio_netio_queue,
+ .setup = fio_netio_setup_splice,
+ .init = fio_netio_init,
+ .cleanup = fio_netio_cleanup,
+ .open_file = fio_netio_open_file,
+ .close_file = generic_close_file,
+ .flags = FIO_SYNCIO | FIO_DISKLESSIO | FIO_UNIDIR |
+ FIO_SIGTERM | FIO_PIPEIO,
+};
+#endif
+
+static struct ioengine_ops ioengine_rw = {
.name = "net",
.version = FIO_IOOPS_VERSION,
.prep = fio_netio_prep,
.init = fio_netio_init,
.cleanup = fio_netio_cleanup,
.open_file = fio_netio_open_file,
- .close_file = generic_close_file,
- .flags = FIO_SYNCIO | FIO_DISKLESSIO,
+ .close_file = fio_netio_close_file,
+ .flags = FIO_SYNCIO | FIO_DISKLESSIO | FIO_UNIDIR |
+ FIO_SIGTERM | FIO_PIPEIO,
};
static void fio_init fio_netio_register(void)
{
- register_ioengine(&ioengine);
+ register_ioengine(&ioengine_rw);
+#ifdef FIO_HAVE_SPLICE
+ register_ioengine(&ioengine_splice);
+#endif
}
static void fio_exit fio_netio_unregister(void)
{
- unregister_ioengine(&ioengine);
+ unregister_ioengine(&ioengine_rw);
+#ifdef FIO_HAVE_SPLICE
+ unregister_ioengine(&ioengine_splice);
+#endif
}
projects / fio.git / blobdiff