2015-02-27 17:19:33 +01:00
|
|
|
/*
|
|
|
|
* QEMU I/O channels sockets driver
|
|
|
|
*
|
|
|
|
* Copyright (c) 2015 Red Hat, Inc.
|
|
|
|
*
|
|
|
|
* This library is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
|
|
* License as published by the Free Software Foundation; either
|
2020-10-14 15:40:33 +02:00
|
|
|
* version 2.1 of the License, or (at your option) any later version.
|
2015-02-27 17:19:33 +01:00
|
|
|
*
|
|
|
|
* This library 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
|
|
|
|
* Lesser General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
|
|
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
|
|
|
|
*/
|
|
|
|
|
2016-01-29 18:50:03 +01:00
|
|
|
#include "qemu/osdep.h"
|
include/qemu/osdep.h: Don't include qapi/error.h
Commit 57cb38b included qapi/error.h into qemu/osdep.h to get the
Error typedef. Since then, we've moved to include qemu/osdep.h
everywhere. Its file comment explains: "To avoid getting into
possible circular include dependencies, this file should not include
any other QEMU headers, with the exceptions of config-host.h,
compiler.h, os-posix.h and os-win32.h, all of which are doing a
similar job to this file and are under similar constraints."
qapi/error.h doesn't do a similar job, and it doesn't adhere to
similar constraints: it includes qapi-types.h. That's in excess of
100KiB of crap most .c files don't actually need.
Add the typedef to qemu/typedefs.h, and include that instead of
qapi/error.h. Include qapi/error.h in .c files that need it and don't
get it now. Include qapi-types.h in qom/object.h for uint16List.
Update scripts/clean-includes accordingly. Update it further to match
reality: replace config.h by config-target.h, add sysemu/os-posix.h,
sysemu/os-win32.h. Update the list of includes in the qemu/osdep.h
comment quoted above similarly.
This reduces the number of objects depending on qapi/error.h from "all
of them" to less than a third. Unfortunately, the number depending on
qapi-types.h shrinks only a little. More work is needed for that one.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[Fix compilation without the spice devel packages. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-14 09:01:28 +01:00
|
|
|
#include "qapi/error.h"
|
2018-02-11 10:36:01 +01:00
|
|
|
#include "qapi/qapi-visit-sockets.h"
|
2019-05-23 16:35:07 +02:00
|
|
|
#include "qemu/module.h"
|
2015-02-27 17:19:33 +01:00
|
|
|
#include "io/channel-socket.h"
|
|
|
|
#include "io/channel-watch.h"
|
|
|
|
#include "trace.h"
|
2016-06-09 18:48:45 +02:00
|
|
|
#include "qapi/clone-visitor.h"
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
#ifdef CONFIG_LINUX
|
|
|
|
#include <linux/errqueue.h>
|
|
|
|
#include <sys/socket.h>
|
|
|
|
|
|
|
|
#if (defined(MSG_ZEROCOPY) && defined(SO_ZEROCOPY))
|
|
|
|
#define QEMU_MSG_ZEROCOPY
|
|
|
|
#endif
|
|
|
|
#endif
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
#define SOCKET_MAX_FDS 16
|
|
|
|
|
|
|
|
SocketAddress *
|
|
|
|
qio_channel_socket_get_local_address(QIOChannelSocket *ioc,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
return socket_sockaddr_to_address(&ioc->localAddr,
|
|
|
|
ioc->localAddrLen,
|
|
|
|
errp);
|
|
|
|
}
|
|
|
|
|
|
|
|
SocketAddress *
|
|
|
|
qio_channel_socket_get_remote_address(QIOChannelSocket *ioc,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
return socket_sockaddr_to_address(&ioc->remoteAddr,
|
|
|
|
ioc->remoteAddrLen,
|
|
|
|
errp);
|
|
|
|
}
|
|
|
|
|
|
|
|
QIOChannelSocket *
|
|
|
|
qio_channel_socket_new(void)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc;
|
|
|
|
QIOChannel *ioc;
|
|
|
|
|
|
|
|
sioc = QIO_CHANNEL_SOCKET(object_new(TYPE_QIO_CHANNEL_SOCKET));
|
|
|
|
sioc->fd = -1;
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
sioc->zero_copy_queued = 0;
|
|
|
|
sioc->zero_copy_sent = 0;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
ioc = QIO_CHANNEL(sioc);
|
2016-09-29 17:52:37 +02:00
|
|
|
qio_channel_set_feature(ioc, QIO_CHANNEL_FEATURE_SHUTDOWN);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2016-03-07 12:12:36 +01:00
|
|
|
#ifdef WIN32
|
|
|
|
ioc->event = CreateEvent(NULL, FALSE, FALSE, NULL);
|
|
|
|
#endif
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
trace_qio_channel_socket_new(sioc);
|
|
|
|
|
|
|
|
return sioc;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
qio_channel_socket_set_fd(QIOChannelSocket *sioc,
|
|
|
|
int fd,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
if (sioc->fd != -1) {
|
|
|
|
error_setg(errp, "Socket is already open");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
sioc->fd = fd;
|
|
|
|
sioc->remoteAddrLen = sizeof(sioc->remoteAddr);
|
|
|
|
sioc->localAddrLen = sizeof(sioc->localAddr);
|
|
|
|
|
|
|
|
|
|
|
|
if (getpeername(fd, (struct sockaddr *)&sioc->remoteAddr,
|
|
|
|
&sioc->remoteAddrLen) < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
if (errno == ENOTCONN) {
|
2015-02-27 17:19:33 +01:00
|
|
|
memset(&sioc->remoteAddr, 0, sizeof(sioc->remoteAddr));
|
|
|
|
sioc->remoteAddrLen = sizeof(sioc->remoteAddr);
|
|
|
|
} else {
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Unable to query remote socket address");
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (getsockname(fd, (struct sockaddr *)&sioc->localAddr,
|
|
|
|
&sioc->localAddrLen) < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Unable to query local socket address");
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifndef WIN32
|
|
|
|
if (sioc->localAddr.ss_family == AF_UNIX) {
|
|
|
|
QIOChannel *ioc = QIO_CHANNEL(sioc);
|
2016-09-29 17:52:37 +02:00
|
|
|
qio_channel_set_feature(ioc, QIO_CHANNEL_FEATURE_FD_PASS);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
#endif /* WIN32 */
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
error:
|
|
|
|
sioc->fd = -1; /* Let the caller close FD on failure */
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
QIOChannelSocket *
|
|
|
|
qio_channel_socket_new_fd(int fd,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *ioc;
|
|
|
|
|
|
|
|
ioc = qio_channel_socket_new();
|
|
|
|
if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
|
|
|
|
object_unref(OBJECT(ioc));
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
trace_qio_channel_socket_new_fd(ioc, fd);
|
|
|
|
|
|
|
|
return ioc;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
int qio_channel_socket_connect_sync(QIOChannelSocket *ioc,
|
|
|
|
SocketAddress *addr,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
int fd;
|
|
|
|
|
|
|
|
trace_qio_channel_socket_connect_sync(ioc, addr);
|
2017-06-16 10:54:45 +02:00
|
|
|
fd = socket_connect(addr, errp);
|
2015-02-27 17:19:33 +01:00
|
|
|
if (fd < 0) {
|
|
|
|
trace_qio_channel_socket_connect_fail(ioc);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
trace_qio_channel_socket_connect_complete(ioc, fd);
|
|
|
|
if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
|
|
|
|
close(fd);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
#ifdef QEMU_MSG_ZEROCOPY
|
|
|
|
int ret, v = 1;
|
|
|
|
ret = setsockopt(fd, SOL_SOCKET, SO_ZEROCOPY, &v, sizeof(v));
|
|
|
|
if (ret == 0) {
|
|
|
|
/* Zero copy available on host */
|
|
|
|
qio_channel_set_feature(QIO_CHANNEL(ioc),
|
|
|
|
QIO_CHANNEL_FEATURE_WRITE_ZERO_COPY);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2016-08-11 18:38:07 +02:00
|
|
|
static void qio_channel_socket_connect_worker(QIOTask *task,
|
|
|
|
gpointer opaque)
|
2015-02-27 17:19:33 +01:00
|
|
|
{
|
|
|
|
QIOChannelSocket *ioc = QIO_CHANNEL_SOCKET(qio_task_get_source(task));
|
|
|
|
SocketAddress *addr = opaque;
|
2016-08-11 18:38:07 +02:00
|
|
|
Error *err = NULL;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2016-08-11 18:38:07 +02:00
|
|
|
qio_channel_socket_connect_sync(ioc, addr, &err);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2016-08-11 18:38:07 +02:00
|
|
|
qio_task_set_error(task, err);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void qio_channel_socket_connect_async(QIOChannelSocket *ioc,
|
|
|
|
SocketAddress *addr,
|
|
|
|
QIOTaskFunc callback,
|
|
|
|
gpointer opaque,
|
2018-03-05 07:43:23 +01:00
|
|
|
GDestroyNotify destroy,
|
|
|
|
GMainContext *context)
|
2015-02-27 17:19:33 +01:00
|
|
|
{
|
|
|
|
QIOTask *task = qio_task_new(
|
|
|
|
OBJECT(ioc), callback, opaque, destroy);
|
|
|
|
SocketAddress *addrCopy;
|
|
|
|
|
2016-06-09 18:48:45 +02:00
|
|
|
addrCopy = QAPI_CLONE(SocketAddress, addr);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
/* socket_connect() does a non-blocking connect(), but it
|
|
|
|
* still blocks in DNS lookups, so we must use a thread */
|
|
|
|
trace_qio_channel_socket_connect_async(ioc, addr);
|
|
|
|
qio_task_run_in_thread(task,
|
|
|
|
qio_channel_socket_connect_worker,
|
|
|
|
addrCopy,
|
2018-03-05 07:43:22 +01:00
|
|
|
(GDestroyNotify)qapi_free_SocketAddress,
|
2018-03-05 07:43:23 +01:00
|
|
|
context);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
int qio_channel_socket_listen_sync(QIOChannelSocket *ioc,
|
|
|
|
SocketAddress *addr,
|
2019-08-19 15:29:58 +02:00
|
|
|
int num,
|
2015-02-27 17:19:33 +01:00
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
int fd;
|
|
|
|
|
2019-08-19 15:29:58 +02:00
|
|
|
trace_qio_channel_socket_listen_sync(ioc, addr, num);
|
|
|
|
fd = socket_listen(addr, num, errp);
|
2015-02-27 17:19:33 +01:00
|
|
|
if (fd < 0) {
|
|
|
|
trace_qio_channel_socket_listen_fail(ioc);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
trace_qio_channel_socket_listen_complete(ioc, fd);
|
|
|
|
if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
|
|
|
|
close(fd);
|
|
|
|
return -1;
|
|
|
|
}
|
2016-10-26 18:20:20 +02:00
|
|
|
qio_channel_set_feature(QIO_CHANNEL(ioc), QIO_CHANNEL_FEATURE_LISTEN);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2019-08-20 09:40:39 +02:00
|
|
|
struct QIOChannelListenWorkerData {
|
|
|
|
SocketAddress *addr;
|
|
|
|
int num; /* amount of expected connections */
|
|
|
|
};
|
|
|
|
|
|
|
|
static void qio_channel_listen_worker_free(gpointer opaque)
|
|
|
|
{
|
|
|
|
struct QIOChannelListenWorkerData *data = opaque;
|
|
|
|
|
|
|
|
qapi_free_SocketAddress(data->addr);
|
|
|
|
g_free(data);
|
|
|
|
}
|
|
|
|
|
2016-08-11 18:38:07 +02:00
|
|
|
static void qio_channel_socket_listen_worker(QIOTask *task,
|
|
|
|
gpointer opaque)
|
2015-02-27 17:19:33 +01:00
|
|
|
{
|
|
|
|
QIOChannelSocket *ioc = QIO_CHANNEL_SOCKET(qio_task_get_source(task));
|
2019-08-20 09:40:39 +02:00
|
|
|
struct QIOChannelListenWorkerData *data = opaque;
|
2016-08-11 18:38:07 +02:00
|
|
|
Error *err = NULL;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2019-08-20 09:40:39 +02:00
|
|
|
qio_channel_socket_listen_sync(ioc, data->addr, data->num, &err);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2016-08-11 18:38:07 +02:00
|
|
|
qio_task_set_error(task, err);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void qio_channel_socket_listen_async(QIOChannelSocket *ioc,
|
|
|
|
SocketAddress *addr,
|
2019-08-20 09:40:39 +02:00
|
|
|
int num,
|
2015-02-27 17:19:33 +01:00
|
|
|
QIOTaskFunc callback,
|
|
|
|
gpointer opaque,
|
2018-03-05 07:43:23 +01:00
|
|
|
GDestroyNotify destroy,
|
|
|
|
GMainContext *context)
|
2015-02-27 17:19:33 +01:00
|
|
|
{
|
|
|
|
QIOTask *task = qio_task_new(
|
|
|
|
OBJECT(ioc), callback, opaque, destroy);
|
2019-08-20 09:40:39 +02:00
|
|
|
struct QIOChannelListenWorkerData *data;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2019-08-20 09:40:39 +02:00
|
|
|
data = g_new0(struct QIOChannelListenWorkerData, 1);
|
|
|
|
data->addr = QAPI_CLONE(SocketAddress, addr);
|
|
|
|
data->num = num;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
/* socket_listen() blocks in DNS lookups, so we must use a thread */
|
2019-08-20 09:40:39 +02:00
|
|
|
trace_qio_channel_socket_listen_async(ioc, addr, num);
|
2015-02-27 17:19:33 +01:00
|
|
|
qio_task_run_in_thread(task,
|
|
|
|
qio_channel_socket_listen_worker,
|
2019-08-20 09:40:39 +02:00
|
|
|
data,
|
|
|
|
qio_channel_listen_worker_free,
|
2018-03-05 07:43:23 +01:00
|
|
|
context);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
int qio_channel_socket_dgram_sync(QIOChannelSocket *ioc,
|
|
|
|
SocketAddress *localAddr,
|
|
|
|
SocketAddress *remoteAddr,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
int fd;
|
|
|
|
|
|
|
|
trace_qio_channel_socket_dgram_sync(ioc, localAddr, remoteAddr);
|
2016-02-09 11:59:15 +01:00
|
|
|
fd = socket_dgram(remoteAddr, localAddr, errp);
|
2015-02-27 17:19:33 +01:00
|
|
|
if (fd < 0) {
|
|
|
|
trace_qio_channel_socket_dgram_fail(ioc);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
trace_qio_channel_socket_dgram_complete(ioc, fd);
|
|
|
|
if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
|
|
|
|
close(fd);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
struct QIOChannelSocketDGramWorkerData {
|
|
|
|
SocketAddress *localAddr;
|
|
|
|
SocketAddress *remoteAddr;
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
static void qio_channel_socket_dgram_worker_free(gpointer opaque)
|
|
|
|
{
|
|
|
|
struct QIOChannelSocketDGramWorkerData *data = opaque;
|
|
|
|
qapi_free_SocketAddress(data->localAddr);
|
|
|
|
qapi_free_SocketAddress(data->remoteAddr);
|
|
|
|
g_free(data);
|
|
|
|
}
|
|
|
|
|
2016-08-11 18:38:07 +02:00
|
|
|
static void qio_channel_socket_dgram_worker(QIOTask *task,
|
|
|
|
gpointer opaque)
|
2015-02-27 17:19:33 +01:00
|
|
|
{
|
|
|
|
QIOChannelSocket *ioc = QIO_CHANNEL_SOCKET(qio_task_get_source(task));
|
|
|
|
struct QIOChannelSocketDGramWorkerData *data = opaque;
|
2016-08-11 18:38:07 +02:00
|
|
|
Error *err = NULL;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
/* socket_dgram() blocks in DNS lookups, so we must use a thread */
|
2016-08-11 18:38:07 +02:00
|
|
|
qio_channel_socket_dgram_sync(ioc, data->localAddr,
|
|
|
|
data->remoteAddr, &err);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2016-08-11 18:38:07 +02:00
|
|
|
qio_task_set_error(task, err);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void qio_channel_socket_dgram_async(QIOChannelSocket *ioc,
|
|
|
|
SocketAddress *localAddr,
|
|
|
|
SocketAddress *remoteAddr,
|
|
|
|
QIOTaskFunc callback,
|
|
|
|
gpointer opaque,
|
2018-03-05 07:43:23 +01:00
|
|
|
GDestroyNotify destroy,
|
|
|
|
GMainContext *context)
|
2015-02-27 17:19:33 +01:00
|
|
|
{
|
|
|
|
QIOTask *task = qio_task_new(
|
|
|
|
OBJECT(ioc), callback, opaque, destroy);
|
|
|
|
struct QIOChannelSocketDGramWorkerData *data = g_new0(
|
|
|
|
struct QIOChannelSocketDGramWorkerData, 1);
|
|
|
|
|
2016-06-09 18:48:45 +02:00
|
|
|
data->localAddr = QAPI_CLONE(SocketAddress, localAddr);
|
|
|
|
data->remoteAddr = QAPI_CLONE(SocketAddress, remoteAddr);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
trace_qio_channel_socket_dgram_async(ioc, localAddr, remoteAddr);
|
|
|
|
qio_task_run_in_thread(task,
|
|
|
|
qio_channel_socket_dgram_worker,
|
|
|
|
data,
|
2018-03-05 07:43:22 +01:00
|
|
|
qio_channel_socket_dgram_worker_free,
|
2018-03-05 07:43:23 +01:00
|
|
|
context);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
QIOChannelSocket *
|
|
|
|
qio_channel_socket_accept(QIOChannelSocket *ioc,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *cioc;
|
|
|
|
|
2017-04-03 13:05:21 +02:00
|
|
|
cioc = qio_channel_socket_new();
|
2015-02-27 17:19:33 +01:00
|
|
|
cioc->remoteAddrLen = sizeof(ioc->remoteAddr);
|
|
|
|
cioc->localAddrLen = sizeof(ioc->localAddr);
|
|
|
|
|
|
|
|
retry:
|
|
|
|
trace_qio_channel_socket_accept(ioc);
|
2016-03-10 18:07:27 +01:00
|
|
|
cioc->fd = qemu_accept(ioc->fd, (struct sockaddr *)&cioc->remoteAddr,
|
|
|
|
&cioc->remoteAddrLen);
|
2015-02-27 17:19:33 +01:00
|
|
|
if (cioc->fd < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
if (errno == EINTR) {
|
2015-02-27 17:19:33 +01:00
|
|
|
goto retry;
|
|
|
|
}
|
2017-08-02 11:41:20 +02:00
|
|
|
error_setg_errno(errp, errno, "Unable to accept connection");
|
|
|
|
trace_qio_channel_socket_accept_fail(ioc);
|
2015-02-27 17:19:33 +01:00
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2015-12-21 13:04:21 +01:00
|
|
|
if (getsockname(cioc->fd, (struct sockaddr *)&cioc->localAddr,
|
|
|
|
&cioc->localAddrLen) < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Unable to query local socket address");
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2015-12-21 13:04:21 +01:00
|
|
|
#ifndef WIN32
|
|
|
|
if (cioc->localAddr.ss_family == AF_UNIX) {
|
2016-09-29 17:52:37 +02:00
|
|
|
QIOChannel *ioc_local = QIO_CHANNEL(cioc);
|
|
|
|
qio_channel_set_feature(ioc_local, QIO_CHANNEL_FEATURE_FD_PASS);
|
2015-12-21 13:04:21 +01:00
|
|
|
}
|
|
|
|
#endif /* WIN32 */
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
trace_qio_channel_socket_accept_complete(ioc, cioc, cioc->fd);
|
|
|
|
return cioc;
|
|
|
|
|
|
|
|
error:
|
|
|
|
object_unref(OBJECT(cioc));
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void qio_channel_socket_init(Object *obj)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *ioc = QIO_CHANNEL_SOCKET(obj);
|
|
|
|
ioc->fd = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void qio_channel_socket_finalize(Object *obj)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *ioc = QIO_CHANNEL_SOCKET(obj);
|
2016-06-16 21:28:52 +02:00
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
if (ioc->fd != -1) {
|
2016-09-29 17:52:36 +02:00
|
|
|
QIOChannel *ioc_local = QIO_CHANNEL(ioc);
|
|
|
|
if (qio_channel_has_feature(ioc_local, QIO_CHANNEL_FEATURE_LISTEN)) {
|
2016-06-16 21:28:52 +02:00
|
|
|
Error *err = NULL;
|
|
|
|
|
|
|
|
socket_listen_cleanup(ioc->fd, &err);
|
|
|
|
if (err) {
|
|
|
|
error_report_err(err);
|
|
|
|
err = NULL;
|
|
|
|
}
|
|
|
|
}
|
2016-03-07 12:12:36 +01:00
|
|
|
#ifdef WIN32
|
|
|
|
WSAEventSelect(ioc->fd, NULL, 0);
|
|
|
|
#endif
|
|
|
|
closesocket(ioc->fd);
|
2015-02-27 17:19:33 +01:00
|
|
|
ioc->fd = -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
#ifndef WIN32
|
|
|
|
static void qio_channel_socket_copy_fds(struct msghdr *msg,
|
|
|
|
int **fds, size_t *nfds)
|
|
|
|
{
|
|
|
|
struct cmsghdr *cmsg;
|
|
|
|
|
|
|
|
*nfds = 0;
|
|
|
|
*fds = NULL;
|
|
|
|
|
|
|
|
for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
|
|
|
|
int fd_size, i;
|
|
|
|
int gotfds;
|
|
|
|
|
|
|
|
if (cmsg->cmsg_len < CMSG_LEN(sizeof(int)) ||
|
|
|
|
cmsg->cmsg_level != SOL_SOCKET ||
|
|
|
|
cmsg->cmsg_type != SCM_RIGHTS) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
fd_size = cmsg->cmsg_len - CMSG_LEN(0);
|
|
|
|
|
|
|
|
if (!fd_size) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
gotfds = fd_size / sizeof(int);
|
|
|
|
*fds = g_renew(int, *fds, *nfds + gotfds);
|
|
|
|
memcpy(*fds + *nfds, CMSG_DATA(cmsg), fd_size);
|
|
|
|
|
|
|
|
for (i = 0; i < gotfds; i++) {
|
|
|
|
int fd = (*fds)[*nfds + i];
|
|
|
|
if (fd < 0) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* O_NONBLOCK is preserved across SCM_RIGHTS so reset it */
|
2022-04-25 15:33:47 +02:00
|
|
|
qemu_socket_set_block(fd);
|
2015-02-27 17:19:33 +01:00
|
|
|
|
|
|
|
#ifndef MSG_CMSG_CLOEXEC
|
|
|
|
qemu_set_cloexec(fd);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
*nfds += gotfds;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static ssize_t qio_channel_socket_readv(QIOChannel *ioc,
|
|
|
|
const struct iovec *iov,
|
|
|
|
size_t niov,
|
|
|
|
int **fds,
|
|
|
|
size_t *nfds,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
ssize_t ret;
|
|
|
|
struct msghdr msg = { NULL, };
|
|
|
|
char control[CMSG_SPACE(sizeof(int) * SOCKET_MAX_FDS)];
|
|
|
|
int sflags = 0;
|
|
|
|
|
2016-01-18 11:37:21 +01:00
|
|
|
memset(control, 0, CMSG_SPACE(sizeof(int) * SOCKET_MAX_FDS));
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
msg.msg_iov = (struct iovec *)iov;
|
|
|
|
msg.msg_iovlen = niov;
|
|
|
|
if (fds && nfds) {
|
|
|
|
msg.msg_control = control;
|
|
|
|
msg.msg_controllen = sizeof(control);
|
2021-04-21 13:28:30 +02:00
|
|
|
#ifdef MSG_CMSG_CLOEXEC
|
|
|
|
sflags |= MSG_CMSG_CLOEXEC;
|
|
|
|
#endif
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
retry:
|
|
|
|
ret = recvmsg(sioc->fd, &msg, sflags);
|
|
|
|
if (ret < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
if (errno == EAGAIN) {
|
2015-02-27 17:19:33 +01:00
|
|
|
return QIO_CHANNEL_ERR_BLOCK;
|
|
|
|
}
|
2016-03-07 21:36:03 +01:00
|
|
|
if (errno == EINTR) {
|
2015-02-27 17:19:33 +01:00
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Unable to read from socket");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (fds && nfds) {
|
|
|
|
qio_channel_socket_copy_fds(&msg, fds, nfds);
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t qio_channel_socket_writev(QIOChannel *ioc,
|
|
|
|
const struct iovec *iov,
|
|
|
|
size_t niov,
|
|
|
|
int *fds,
|
|
|
|
size_t nfds,
|
2022-05-13 08:28:31 +02:00
|
|
|
int flags,
|
2015-02-27 17:19:33 +01:00
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
ssize_t ret;
|
|
|
|
struct msghdr msg = { NULL, };
|
2016-01-18 11:37:21 +01:00
|
|
|
char control[CMSG_SPACE(sizeof(int) * SOCKET_MAX_FDS)];
|
2015-12-21 12:58:51 +01:00
|
|
|
size_t fdsize = sizeof(int) * nfds;
|
|
|
|
struct cmsghdr *cmsg;
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
int sflags = 0;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2016-01-18 11:37:21 +01:00
|
|
|
memset(control, 0, CMSG_SPACE(sizeof(int) * SOCKET_MAX_FDS));
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
msg.msg_iov = (struct iovec *)iov;
|
|
|
|
msg.msg_iovlen = niov;
|
|
|
|
|
|
|
|
if (nfds) {
|
|
|
|
if (nfds > SOCKET_MAX_FDS) {
|
2016-01-11 13:59:44 +01:00
|
|
|
error_setg_errno(errp, EINVAL,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Only %d FDs can be sent, got %zu",
|
|
|
|
SOCKET_MAX_FDS, nfds);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
msg.msg_control = control;
|
|
|
|
msg.msg_controllen = CMSG_SPACE(sizeof(int) * nfds);
|
|
|
|
|
|
|
|
cmsg = CMSG_FIRSTHDR(&msg);
|
|
|
|
cmsg->cmsg_len = CMSG_LEN(fdsize);
|
|
|
|
cmsg->cmsg_level = SOL_SOCKET;
|
|
|
|
cmsg->cmsg_type = SCM_RIGHTS;
|
|
|
|
memcpy(CMSG_DATA(cmsg), fds, fdsize);
|
|
|
|
}
|
|
|
|
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
if (flags & QIO_CHANNEL_WRITE_FLAG_ZERO_COPY) {
|
2022-06-20 07:39:42 +02:00
|
|
|
#ifdef QEMU_MSG_ZEROCOPY
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
sflags = MSG_ZEROCOPY;
|
2022-06-20 07:39:42 +02:00
|
|
|
#else
|
|
|
|
/*
|
|
|
|
* We expect QIOChannel class entry point to have
|
|
|
|
* blocked this code path already
|
|
|
|
*/
|
|
|
|
g_assert_not_reached();
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
#endif
|
2022-06-20 07:39:42 +02:00
|
|
|
}
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
retry:
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
ret = sendmsg(sioc->fd, &msg, sflags);
|
2015-02-27 17:19:33 +01:00
|
|
|
if (ret <= 0) {
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
switch (errno) {
|
|
|
|
case EAGAIN:
|
2015-02-27 17:19:33 +01:00
|
|
|
return QIO_CHANNEL_ERR_BLOCK;
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
case EINTR:
|
2015-02-27 17:19:33 +01:00
|
|
|
goto retry;
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
case ENOBUFS:
|
2022-06-20 07:39:42 +02:00
|
|
|
if (flags & QIO_CHANNEL_WRITE_FLAG_ZERO_COPY) {
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
error_setg_errno(errp, errno,
|
|
|
|
"Process can't lock enough memory for using MSG_ZEROCOPY");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
break;
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Unable to write to socket");
|
|
|
|
return -1;
|
|
|
|
}
|
2022-06-20 07:39:43 +02:00
|
|
|
|
|
|
|
if (flags & QIO_CHANNEL_WRITE_FLAG_ZERO_COPY) {
|
|
|
|
sioc->zero_copy_queued++;
|
|
|
|
}
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
#else /* WIN32 */
|
|
|
|
static ssize_t qio_channel_socket_readv(QIOChannel *ioc,
|
|
|
|
const struct iovec *iov,
|
|
|
|
size_t niov,
|
|
|
|
int **fds,
|
|
|
|
size_t *nfds,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
ssize_t done = 0;
|
|
|
|
ssize_t i;
|
|
|
|
|
|
|
|
for (i = 0; i < niov; i++) {
|
|
|
|
ssize_t ret;
|
|
|
|
retry:
|
|
|
|
ret = recv(sioc->fd,
|
|
|
|
iov[i].iov_base,
|
|
|
|
iov[i].iov_len,
|
|
|
|
0);
|
|
|
|
if (ret < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
if (errno == EAGAIN) {
|
2015-02-27 17:19:33 +01:00
|
|
|
if (done) {
|
|
|
|
return done;
|
|
|
|
} else {
|
|
|
|
return QIO_CHANNEL_ERR_BLOCK;
|
|
|
|
}
|
2016-03-07 21:36:03 +01:00
|
|
|
} else if (errno == EINTR) {
|
2015-02-27 17:19:33 +01:00
|
|
|
goto retry;
|
|
|
|
} else {
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2016-03-08 13:06:30 +01:00
|
|
|
"Unable to read from socket");
|
2015-02-27 17:19:33 +01:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
done += ret;
|
|
|
|
if (ret < iov[i].iov_len) {
|
|
|
|
return done;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return done;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t qio_channel_socket_writev(QIOChannel *ioc,
|
|
|
|
const struct iovec *iov,
|
|
|
|
size_t niov,
|
|
|
|
int *fds,
|
|
|
|
size_t nfds,
|
2022-05-13 08:28:31 +02:00
|
|
|
int flags,
|
2015-02-27 17:19:33 +01:00
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
ssize_t done = 0;
|
|
|
|
ssize_t i;
|
|
|
|
|
|
|
|
for (i = 0; i < niov; i++) {
|
|
|
|
ssize_t ret;
|
|
|
|
retry:
|
|
|
|
ret = send(sioc->fd,
|
|
|
|
iov[i].iov_base,
|
|
|
|
iov[i].iov_len,
|
|
|
|
0);
|
|
|
|
if (ret < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
if (errno == EAGAIN) {
|
2015-02-27 17:19:33 +01:00
|
|
|
if (done) {
|
|
|
|
return done;
|
|
|
|
} else {
|
|
|
|
return QIO_CHANNEL_ERR_BLOCK;
|
|
|
|
}
|
2016-03-07 21:36:03 +01:00
|
|
|
} else if (errno == EINTR) {
|
2015-02-27 17:19:33 +01:00
|
|
|
goto retry;
|
|
|
|
} else {
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Unable to write to socket");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
done += ret;
|
|
|
|
if (ret < iov[i].iov_len) {
|
|
|
|
return done;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return done;
|
|
|
|
}
|
|
|
|
#endif /* WIN32 */
|
|
|
|
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
|
|
|
|
#ifdef QEMU_MSG_ZEROCOPY
|
|
|
|
static int qio_channel_socket_flush(QIOChannel *ioc,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
struct msghdr msg = {};
|
|
|
|
struct sock_extended_err *serr;
|
|
|
|
struct cmsghdr *cm;
|
|
|
|
char control[CMSG_SPACE(sizeof(*serr))];
|
|
|
|
int received;
|
2022-07-11 23:11:11 +02:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (sioc->zero_copy_queued == sioc->zero_copy_sent) {
|
|
|
|
return 0;
|
|
|
|
}
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
|
|
|
|
msg.msg_control = control;
|
|
|
|
msg.msg_controllen = sizeof(control);
|
|
|
|
memset(control, 0, sizeof(control));
|
|
|
|
|
2022-07-11 23:11:11 +02:00
|
|
|
ret = 1;
|
|
|
|
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
while (sioc->zero_copy_sent < sioc->zero_copy_queued) {
|
|
|
|
received = recvmsg(sioc->fd, &msg, MSG_ERRQUEUE);
|
|
|
|
if (received < 0) {
|
|
|
|
switch (errno) {
|
|
|
|
case EAGAIN:
|
|
|
|
/* Nothing on errqueue, wait until something is available */
|
|
|
|
qio_channel_wait(ioc, G_IO_ERR);
|
|
|
|
continue;
|
|
|
|
case EINTR:
|
|
|
|
continue;
|
|
|
|
default:
|
|
|
|
error_setg_errno(errp, errno,
|
|
|
|
"Unable to read errqueue");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
cm = CMSG_FIRSTHDR(&msg);
|
2022-08-04 09:10:43 +02:00
|
|
|
if (cm->cmsg_level != SOL_IP && cm->cmsg_type != IP_RECVERR &&
|
|
|
|
cm->cmsg_level != SOL_IPV6 && cm->cmsg_type != IPV6_RECVERR) {
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
error_setg_errno(errp, EPROTOTYPE,
|
|
|
|
"Wrong cmsg in errqueue");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
serr = (void *) CMSG_DATA(cm);
|
|
|
|
if (serr->ee_errno != SO_EE_ORIGIN_NONE) {
|
|
|
|
error_setg_errno(errp, serr->ee_errno,
|
|
|
|
"Error on socket");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
if (serr->ee_origin != SO_EE_ORIGIN_ZEROCOPY) {
|
|
|
|
error_setg_errno(errp, serr->ee_origin,
|
|
|
|
"Error not from zero copy");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* No errors, count successfully finished sendmsg()*/
|
|
|
|
sioc->zero_copy_sent += serr->ee_data - serr->ee_info + 1;
|
|
|
|
|
|
|
|
/* If any sendmsg() succeeded using zero copy, return 0 at the end */
|
|
|
|
if (serr->ee_code != SO_EE_CODE_ZEROCOPY_COPIED) {
|
|
|
|
ret = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* QEMU_MSG_ZEROCOPY */
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
static int
|
|
|
|
qio_channel_socket_set_blocking(QIOChannel *ioc,
|
|
|
|
bool enabled,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
|
|
|
|
if (enabled) {
|
2022-04-25 15:33:47 +02:00
|
|
|
qemu_socket_set_block(sioc->fd);
|
2015-02-27 17:19:33 +01:00
|
|
|
} else {
|
2022-04-25 15:33:47 +02:00
|
|
|
qemu_socket_set_nonblock(sioc->fd);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
qio_channel_socket_set_delay(QIOChannel *ioc,
|
|
|
|
bool enabled)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
int v = enabled ? 0 : 1;
|
|
|
|
|
2022-02-18 22:34:50 +01:00
|
|
|
setsockopt(sioc->fd,
|
|
|
|
IPPROTO_TCP, TCP_NODELAY,
|
|
|
|
&v, sizeof(v));
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
qio_channel_socket_set_cork(QIOChannel *ioc,
|
|
|
|
bool enabled)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
int v = enabled ? 1 : 0;
|
|
|
|
|
|
|
|
socket_set_cork(sioc->fd, v);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
qio_channel_socket_close(QIOChannel *ioc,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
2018-05-21 18:17:35 +02:00
|
|
|
int rc = 0;
|
2020-04-22 15:07:15 +02:00
|
|
|
Error *err = NULL;
|
2015-02-27 17:19:33 +01:00
|
|
|
|
2016-03-07 12:12:36 +01:00
|
|
|
if (sioc->fd != -1) {
|
|
|
|
#ifdef WIN32
|
|
|
|
WSAEventSelect(sioc->fd, NULL, 0);
|
|
|
|
#endif
|
2019-01-14 12:33:18 +01:00
|
|
|
if (qio_channel_has_feature(ioc, QIO_CHANNEL_FEATURE_LISTEN)) {
|
|
|
|
socket_listen_cleanup(sioc->fd, errp);
|
|
|
|
}
|
|
|
|
|
2016-03-07 12:12:36 +01:00
|
|
|
if (closesocket(sioc->fd) < 0) {
|
|
|
|
sioc->fd = -1;
|
2020-04-22 15:07:15 +02:00
|
|
|
error_setg_errno(&err, errno, "Unable to close socket");
|
|
|
|
error_propagate(errp, err);
|
2016-03-07 12:12:36 +01:00
|
|
|
return -1;
|
|
|
|
}
|
2015-02-27 17:19:33 +01:00
|
|
|
sioc->fd = -1;
|
|
|
|
}
|
2018-05-21 18:17:35 +02:00
|
|
|
return rc;
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
qio_channel_socket_shutdown(QIOChannel *ioc,
|
|
|
|
QIOChannelShutdown how,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
|
|
|
int sockhow;
|
|
|
|
|
|
|
|
switch (how) {
|
|
|
|
case QIO_CHANNEL_SHUTDOWN_READ:
|
|
|
|
sockhow = SHUT_RD;
|
|
|
|
break;
|
|
|
|
case QIO_CHANNEL_SHUTDOWN_WRITE:
|
|
|
|
sockhow = SHUT_WR;
|
|
|
|
break;
|
|
|
|
case QIO_CHANNEL_SHUTDOWN_BOTH:
|
|
|
|
default:
|
|
|
|
sockhow = SHUT_RDWR;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (shutdown(sioc->fd, sockhow) < 0) {
|
2016-03-07 21:36:03 +01:00
|
|
|
error_setg_errno(errp, errno,
|
2015-02-27 17:19:33 +01:00
|
|
|
"Unable to shutdown socket");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-02-13 14:52:22 +01:00
|
|
|
static void qio_channel_socket_set_aio_fd_handler(QIOChannel *ioc,
|
|
|
|
AioContext *ctx,
|
|
|
|
IOHandler *io_read,
|
|
|
|
IOHandler *io_write,
|
|
|
|
void *opaque)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
aio-posix: split poll check from ready handler
Adaptive polling measures the execution time of the polling check plus
handlers called when a polled event becomes ready. Handlers can take a
significant amount of time, making it look like polling was running for
a long time when in fact the event handler was running for a long time.
For example, on Linux the io_submit(2) syscall invoked when a virtio-blk
device's virtqueue becomes ready can take 10s of microseconds. This
can exceed the default polling interval (32 microseconds) and cause
adaptive polling to stop polling.
By excluding the handler's execution time from the polling check we make
the adaptive polling calculation more accurate. As a result, the event
loop now stays in polling mode where previously it would have fallen
back to file descriptor monitoring.
The following data was collected with virtio-blk num-queues=2
event_idx=off using an IOThread. Before:
168k IOPS, IOThread syscalls:
9837.115 ( 0.020 ms): IO iothread1/620155 io_submit(ctx_id: 140512552468480, nr: 16, iocbpp: 0x7fcb9f937db0) = 16
9837.158 ( 0.002 ms): IO iothread1/620155 write(fd: 103, buf: 0x556a2ef71b88, count: 8) = 8
9837.161 ( 0.001 ms): IO iothread1/620155 write(fd: 104, buf: 0x556a2ef71b88, count: 8) = 8
9837.163 ( 0.001 ms): IO iothread1/620155 ppoll(ufds: 0x7fcb90002800, nfds: 4, tsp: 0x7fcb9f1342d0, sigsetsize: 8) = 3
9837.164 ( 0.001 ms): IO iothread1/620155 read(fd: 107, buf: 0x7fcb9f939cc0, count: 512) = 8
9837.174 ( 0.001 ms): IO iothread1/620155 read(fd: 105, buf: 0x7fcb9f939cc0, count: 512) = 8
9837.176 ( 0.001 ms): IO iothread1/620155 read(fd: 106, buf: 0x7fcb9f939cc0, count: 512) = 8
9837.209 ( 0.035 ms): IO iothread1/620155 io_submit(ctx_id: 140512552468480, nr: 32, iocbpp: 0x7fca7d0cebe0) = 32
174k IOPS (+3.6%), IOThread syscalls:
9809.566 ( 0.036 ms): IO iothread1/623061 io_submit(ctx_id: 140539805028352, nr: 32, iocbpp: 0x7fd0cdd62be0) = 32
9809.625 ( 0.001 ms): IO iothread1/623061 write(fd: 103, buf: 0x5647cfba5f58, count: 8) = 8
9809.627 ( 0.002 ms): IO iothread1/623061 write(fd: 104, buf: 0x5647cfba5f58, count: 8) = 8
9809.663 ( 0.036 ms): IO iothread1/623061 io_submit(ctx_id: 140539805028352, nr: 32, iocbpp: 0x7fd0d0388b50) = 32
Notice that ppoll(2) and eventfd read(2) syscalls are eliminated because
the IOThread stays in polling mode instead of falling back to file
descriptor monitoring.
As usual, polling is not implemented on Windows so this patch ignores
the new io_poll_read() callback in aio-win32.c.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20211207132336.36627-2-stefanha@redhat.com
[Fixed up aio_set_event_notifier() calls in
tests/unit/test-fdmon-epoll.c added after this series was queued.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2021-12-07 14:23:31 +01:00
|
|
|
aio_set_fd_handler(ctx, sioc->fd, false,
|
|
|
|
io_read, io_write, NULL, NULL, opaque);
|
2017-02-13 14:52:22 +01:00
|
|
|
}
|
|
|
|
|
2015-02-27 17:19:33 +01:00
|
|
|
static GSource *qio_channel_socket_create_watch(QIOChannel *ioc,
|
|
|
|
GIOCondition condition)
|
|
|
|
{
|
|
|
|
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
|
2016-03-07 11:16:39 +01:00
|
|
|
return qio_channel_create_socket_watch(ioc,
|
|
|
|
sioc->fd,
|
|
|
|
condition);
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void qio_channel_socket_class_init(ObjectClass *klass,
|
|
|
|
void *class_data G_GNUC_UNUSED)
|
|
|
|
{
|
|
|
|
QIOChannelClass *ioc_klass = QIO_CHANNEL_CLASS(klass);
|
|
|
|
|
|
|
|
ioc_klass->io_writev = qio_channel_socket_writev;
|
|
|
|
ioc_klass->io_readv = qio_channel_socket_readv;
|
|
|
|
ioc_klass->io_set_blocking = qio_channel_socket_set_blocking;
|
|
|
|
ioc_klass->io_close = qio_channel_socket_close;
|
|
|
|
ioc_klass->io_shutdown = qio_channel_socket_shutdown;
|
|
|
|
ioc_klass->io_set_cork = qio_channel_socket_set_cork;
|
|
|
|
ioc_klass->io_set_delay = qio_channel_socket_set_delay;
|
|
|
|
ioc_klass->io_create_watch = qio_channel_socket_create_watch;
|
2017-02-13 14:52:22 +01:00
|
|
|
ioc_klass->io_set_aio_fd_handler = qio_channel_socket_set_aio_fd_handler;
|
QIOChannelSocket: Implement io_writev zero copy flag & io_flush for CONFIG_LINUX
For CONFIG_LINUX, implement the new zero copy flag and the optional callback
io_flush on QIOChannelSocket, but enables it only when MSG_ZEROCOPY
feature is available in the host kernel, which is checked on
qio_channel_socket_connect_sync()
qio_channel_socket_flush() was implemented by counting how many times
sendmsg(...,MSG_ZEROCOPY) was successfully called, and then reading the
socket's error queue, in order to find how many of them finished sending.
Flush will loop until those counters are the same, or until some error occurs.
Notes on using writev() with QIO_CHANNEL_WRITE_FLAG_ZERO_COPY:
1: Buffer
- As MSG_ZEROCOPY tells the kernel to use the same user buffer to avoid copying,
some caution is necessary to avoid overwriting any buffer before it's sent.
If something like this happen, a newer version of the buffer may be sent instead.
- If this is a problem, it's recommended to call qio_channel_flush() before freeing
or re-using the buffer.
2: Locked memory
- When using MSG_ZERCOCOPY, the buffer memory will be locked after queued, and
unlocked after it's sent.
- Depending on the size of each buffer, and how often it's sent, it may require
a larger amount of locked memory than usually available to non-root user.
- If the required amount of locked memory is not available, writev_zero_copy
will return an error, which can abort an operation like migration,
- Because of this, when an user code wants to add zero copy as a feature, it
requires a mechanism to disable it, so it can still be accessible to less
privileged users.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220513062836.965425-4-leobras@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2022-05-13 08:28:32 +02:00
|
|
|
#ifdef QEMU_MSG_ZEROCOPY
|
|
|
|
ioc_klass->io_flush = qio_channel_socket_flush;
|
|
|
|
#endif
|
2015-02-27 17:19:33 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static const TypeInfo qio_channel_socket_info = {
|
|
|
|
.parent = TYPE_QIO_CHANNEL,
|
|
|
|
.name = TYPE_QIO_CHANNEL_SOCKET,
|
|
|
|
.instance_size = sizeof(QIOChannelSocket),
|
|
|
|
.instance_init = qio_channel_socket_init,
|
|
|
|
.instance_finalize = qio_channel_socket_finalize,
|
|
|
|
.class_init = qio_channel_socket_class_init,
|
|
|
|
};
|
|
|
|
|
|
|
|
static void qio_channel_socket_register_types(void)
|
|
|
|
{
|
|
|
|
type_register_static(&qio_channel_socket_info);
|
|
|
|
}
|
|
|
|
|
|
|
|
type_init(qio_channel_socket_register_types);
|