* [Qemu-devel] Revised Unified Datagram Socket Transport patchset
@ 2017-07-19 20:02 anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 1/4] Unified Datagram Socket Transports anton.ivanov
` (3 more replies)
0 siblings, 4 replies; 5+ messages in thread
From: anton.ivanov @ 2017-07-19 20:02 UTC (permalink / raw)
To: qemu-devel; +Cc: jasowang
Hi Jason, hi list,
Follows a revised patchset. I have addressed most comments.
TODO: replace memcpy with dup where applicable
TODO: add force v4 option
TODO: port the UDP portion of the existing socket transport
to the new infrastructure
Future: add sendmmsg once a "bulk xmit" has been arranged
on the QEMU hw and/or lower network subsystem layers side.
^ permalink raw reply [flat|nested] 5+ messages in thread
* [Qemu-devel] [PATCH 1/4] Unified Datagram Socket Transports
2017-07-19 20:02 [Qemu-devel] Revised Unified Datagram Socket Transport patchset anton.ivanov
@ 2017-07-19 20:02 ` anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 2/4] Migrate l2tpv3 to UDST Backend anton.ivanov
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: anton.ivanov @ 2017-07-19 20:02 UTC (permalink / raw)
To: qemu-devel; +Cc: jasowang, Anton Ivanov
From: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Basic infrastructure to start moving datagram based transports
to a common infrastructure as well as introduce several
additional transports.
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
---
configure | 12 +-
net/Makefile.objs | 2 +-
net/net.c | 4 +-
net/udst.c | 420 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
net/udst.h | 121 ++++++++++++++++
qapi-schema.json | 19 ++-
qemu-options.hx | 2 +-
7 files changed, 569 insertions(+), 11 deletions(-)
create mode 100644 net/udst.c
create mode 100644 net/udst.h
diff --git a/configure b/configure
index bad50f5368..00c911c49b 100755
--- a/configure
+++ b/configure
@@ -1862,7 +1862,9 @@ if ! compile_object -Werror ; then
fi
##########################################
-# L2TPV3 probe
+# UDST probe
+# identical to L2TPv3 probe used for both
+# during migration of L2TPv3 to udst backend
cat > $TMPC <<EOF
#include <sys/socket.h>
@@ -1870,9 +1872,9 @@ cat > $TMPC <<EOF
int main(void) { return sizeof(struct mmsghdr); }
EOF
if compile_prog "" "" ; then
- l2tpv3=yes
+ udst=yes
else
- l2tpv3=no
+ udst=no
fi
##########################################
@@ -5491,8 +5493,8 @@ fi
if test "$netmap" = "yes" ; then
echo "CONFIG_NETMAP=y" >> $config_host_mak
fi
-if test "$l2tpv3" = "yes" ; then
- echo "CONFIG_L2TPV3=y" >> $config_host_mak
+if test "$udst" = "yes" ; then
+ echo "CONFIG_UDST=y" >> $config_host_mak
fi
if test "$cap_ng" = "yes" ; then
echo "CONFIG_LIBCAP=y" >> $config_host_mak
diff --git a/net/Makefile.objs b/net/Makefile.objs
index 67ba5e26fb..ffdfb96bd0 100644
--- a/net/Makefile.objs
+++ b/net/Makefile.objs
@@ -2,7 +2,7 @@ common-obj-y = net.o queue.o checksum.o util.o hub.o
common-obj-y += socket.o
common-obj-y += dump.o
common-obj-y += eth.o
-common-obj-$(CONFIG_L2TPV3) += l2tpv3.o
+common-obj-$(CONFIG_UDST) += udst.o l2tpv3.o
common-obj-$(CONFIG_POSIX) += vhost-user.o
common-obj-$(CONFIG_SLIRP) += slirp.o
common-obj-$(CONFIG_VDE) += vde.o
diff --git a/net/net.c b/net/net.c
index 0e28099554..723a256260 100644
--- a/net/net.c
+++ b/net/net.c
@@ -960,8 +960,8 @@ static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
#ifdef CONFIG_VHOST_NET_USED
[NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
#endif
-#ifdef CONFIG_L2TPV3
- [NET_CLIENT_DRIVER_L2TPV3] = net_init_l2tpv3,
+#ifdef CONFIG_UDST
+ [NET_CLIENT_DRIVER_L2TPV3] = net_init_l2tpv3,
#endif
};
diff --git a/net/udst.c b/net/udst.c
new file mode 100644
index 0000000000..612c90cb3a
--- /dev/null
+++ b/net/udst.c
@@ -0,0 +1,420 @@
+/*
+ * QEMU System Emulator
+ *
+ * Copyright (c) 2015-2017 Cambridge Greys Limited
+ * Copyright (c) 2012-2014 Cisco Systems
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+/*
+ * Udst Datagram Socket Transport Backend
+ * This transport is not intended to be initiated directly by an end-user
+ * It is used as a backend for other transports which use recv/sendmmsg
+ * socket functions for RX/TX.
+ */
+
+#include "qemu/osdep.h"
+#include <linux/ip.h>
+#include <netdb.h>
+#include "net/net.h"
+#include "clients.h"
+#include "qemu-common.h"
+#include "qemu/error-report.h"
+#include "qemu/option.h"
+#include "qemu/sockets.h"
+#include "qemu/iov.h"
+#include "qemu/main-loop.h"
+#include "udst.h"
+
+static void net_udst_send(void *opaque);
+static void udst_writable(void *opaque);
+
+static void udst_update_fd_handler(NetUdstState *s)
+{
+ qemu_set_fd_handler(s->fd,
+ s->read_poll ? net_udst_send : NULL,
+ s->write_poll ? udst_writable : NULL,
+ s);
+}
+
+static void udst_read_poll(NetUdstState *s, bool enable)
+{
+ if (s->read_poll != enable) {
+ s->read_poll = enable;
+ udst_update_fd_handler(s);
+ }
+}
+
+static void udst_write_poll(NetUdstState *s, bool enable)
+{
+ if (s->write_poll != enable) {
+ s->write_poll = enable;
+ udst_update_fd_handler(s);
+ }
+}
+
+static void udst_writable(void *opaque)
+{
+ NetUdstState *s = opaque;
+ udst_write_poll(s, false);
+ qemu_flush_queued_packets(&s->nc);
+}
+
+static void udst_send_completed(NetClientState *nc, ssize_t len)
+{
+ NetUdstState *s = DO_UPCAST(NetUdstState, nc, nc);
+ udst_read_poll(s, true);
+}
+
+static void udst_poll(NetClientState *nc, bool enable)
+{
+ NetUdstState *s = DO_UPCAST(NetUdstState, nc, nc);
+ udst_write_poll(s, enable);
+ udst_read_poll(s, enable);
+}
+
+static ssize_t net_udst_receive_dgram_iov(NetClientState *nc,
+ const struct iovec *iov,
+ int iovcnt)
+{
+ NetUdstState *s = DO_UPCAST(NetUdstState, nc, nc);
+
+ struct msghdr message;
+ int ret;
+
+ if (iovcnt > MAX_UNIFIED_IOVCNT - 1) {
+ error_report(
+ "iovec too long %d > %d, change udst.h",
+ iovcnt, MAX_UNIFIED_IOVCNT
+ );
+ return -1;
+ }
+ if (s->offset > 0) {
+ s->form_header(s);
+ memcpy(s->vec + 1, iov, iovcnt * sizeof(struct iovec));
+ s->vec->iov_base = s->header_buf;
+ s->vec->iov_len = s->offset;
+ message.msg_iovlen = iovcnt + 1;
+ } else {
+ memcpy(s->vec, iov, iovcnt * sizeof(struct iovec));
+ message.msg_iovlen = iovcnt;
+ }
+ message.msg_name = s->dgram_dst;
+ message.msg_namelen = s->dst_size;
+ message.msg_iov = s->vec;
+ message.msg_control = NULL;
+ message.msg_controllen = 0;
+ message.msg_flags = 0;
+ do {
+ ret = sendmsg(s->fd, &message, 0);
+ } while ((ret == -1) && (errno == EINTR));
+ if (ret > 0) {
+ ret -= s->offset;
+ } else if (ret == 0) {
+ /* belt and braces - should not occur on DGRAM
+ * we should get an error and never a 0 send
+ */
+ ret = iov_size(iov, iovcnt);
+ } else {
+ /* signal upper layer that socket buffer is full */
+ ret = -errno;
+ if (ret == -EAGAIN || ret == -ENOBUFS) {
+ udst_write_poll(s, true);
+ ret = 0;
+ }
+ }
+ return ret;
+}
+
+static ssize_t net_udst_receive_dgram(NetClientState *nc,
+ const uint8_t *buf,
+ size_t size)
+{
+ NetUdstState *s = DO_UPCAST(NetUdstState, nc, nc);
+
+ struct iovec *vec;
+ struct msghdr message;
+ ssize_t ret = 0;
+
+ vec = s->vec;
+ if (s->offset > 0) {
+ s->form_header(s);
+ vec->iov_base = s->header_buf;
+ vec->iov_len = s->offset;
+ message.msg_iovlen = 2;
+ vec++;
+ } else {
+ message.msg_iovlen = 1;
+ }
+ vec->iov_base = (void *) buf;
+ vec->iov_len = size;
+ message.msg_name = s->dgram_dst;
+ message.msg_namelen = s->dst_size;
+ message.msg_iov = s->vec;
+ message.msg_control = NULL;
+ message.msg_controllen = 0;
+ message.msg_flags = 0;
+ do {
+ ret = sendmsg(s->fd, &message, 0);
+ } while ((ret == -1) && (errno == EINTR));
+ if (ret > 0) {
+ ret -= s->offset;
+ } else if (ret == 0) {
+ /* belt and braces - should not occur on DGRAM
+ * we should get an error and never a 0 send
+ */
+ ret = size;
+ } else {
+ ret = -errno;
+ if (ret == -EAGAIN || ret == -ENOBUFS) {
+ /* signal upper layer that socket buffer is full */
+ udst_write_poll(s, true);
+ ret = 0;
+ }
+ }
+ return ret;
+}
+
+
+static void net_udst_process_queue(NetUdstState *s)
+{
+ int size = 0;
+ struct iovec *vec;
+ bool bad_read;
+ int data_size;
+ struct mmsghdr *msgvec;
+
+ /* go into ring mode only if there is a "pending" tail */
+ if (s->queue_depth > 0) {
+ do {
+ msgvec = s->msgvec + s->queue_tail;
+ if (msgvec->msg_len > 0) {
+ data_size = msgvec->msg_len - s->header_size;
+ vec = msgvec->msg_hdr.msg_iov;
+ if ((data_size > 0) &&
+ (s->verify_header(s, vec->iov_base) == 0)) {
+ if (s->header_size > 0) {
+ vec++;
+ }
+ /* Use the legacy delivery for now, we will
+ * switch to using our own ring as a queueing mechanism
+ * at a later date
+ */
+ size = qemu_send_packet_async(
+ &s->nc,
+ vec->iov_base,
+ data_size,
+ udst_send_completed
+ );
+ if (size == 0) {
+ udst_read_poll(s, false);
+ }
+ bad_read = false;
+ } else {
+ bad_read = true;
+ if (!s->header_mismatch) {
+ /* report error only once */
+ error_report("udst header verification failed");
+ s->header_mismatch = true;
+ }
+ }
+ } else {
+ bad_read = true;
+ }
+ s->queue_tail = (s->queue_tail + 1) % MAX_UNIFIED_MSGCNT;
+ s->queue_depth--;
+ } while (
+ (s->queue_depth > 0) &&
+ qemu_can_send_packet(&s->nc) &&
+ ((size > 0) || bad_read)
+ );
+ }
+}
+
+static void net_udst_send(void *opaque)
+{
+ NetUdstState *s = opaque;
+ int target_count, count;
+ struct mmsghdr *msgvec;
+
+ /* go into ring mode only if there is a "pending" tail */
+
+ if (s->queue_depth) {
+
+ /* The ring buffer we use has variable intake
+ * count of how much we can read varies - adjust accordingly
+ */
+
+ target_count = MAX_UNIFIED_MSGCNT - s->queue_depth;
+
+ /* Ensure we do not overrun the ring when we have
+ * a lot of enqueued packets
+ */
+
+ if (s->queue_head + target_count > MAX_UNIFIED_MSGCNT) {
+ target_count = MAX_UNIFIED_MSGCNT - s->queue_head;
+ }
+ } else {
+
+ /* we do not have any pending packets - we can use
+ * the whole message vector linearly instead of using
+ * it as a ring
+ */
+
+ s->queue_head = 0;
+ s->queue_tail = 0;
+ target_count = MAX_UNIFIED_MSGCNT;
+ }
+
+ msgvec = s->msgvec + s->queue_head;
+ if (target_count > 0) {
+ do {
+ count = recvmmsg(
+ s->fd,
+ msgvec,
+ target_count, MSG_DONTWAIT, NULL);
+ } while ((count == -1) && (errno == EINTR));
+ if (count < 0) {
+ /* Recv error - we still need to flush packets here,
+ * (re)set queue head to current position
+ */
+ count = 0;
+ }
+ s->queue_head = (s->queue_head + count) % MAX_UNIFIED_MSGCNT;
+ s->queue_depth += count;
+ }
+ net_udst_process_queue(s);
+}
+
+static void destroy_vector(struct mmsghdr *msgvec, int count, int iovcount)
+{
+ int i, j;
+ struct iovec *iov;
+ struct mmsghdr *cleanup = msgvec;
+ if (cleanup) {
+ for (i = 0; i < count; i++) {
+ if (cleanup->msg_hdr.msg_iov) {
+ iov = cleanup->msg_hdr.msg_iov;
+ for (j = 0; j < iovcount; j++) {
+ g_free(iov->iov_base);
+ iov++;
+ }
+ g_free(cleanup->msg_hdr.msg_iov);
+ }
+ cleanup++;
+ }
+ g_free(msgvec);
+ }
+}
+
+
+
+static struct mmsghdr *build_udst_vector(NetUdstState *s, int count)
+{
+ int i;
+ struct iovec *iov;
+ struct mmsghdr *msgvec, *result;
+
+ msgvec = g_new(struct mmsghdr, count);
+ result = msgvec;
+ for (i = 0; i < count ; i++) {
+ msgvec->msg_hdr.msg_name = NULL;
+ msgvec->msg_hdr.msg_namelen = 0;
+ iov = g_new(struct iovec, IOVSIZE);
+ msgvec->msg_hdr.msg_iov = iov;
+ if (s->header_size > 0) {
+ iov->iov_base = g_malloc(s->header_size);
+ iov->iov_len = s->header_size;
+ iov++ ;
+ }
+ iov->iov_base = qemu_memalign(BUFFER_ALIGN, BUFFER_SIZE);
+ iov->iov_len = BUFFER_SIZE;
+ msgvec->msg_hdr.msg_iovlen = 2;
+ msgvec->msg_hdr.msg_control = NULL;
+ msgvec->msg_hdr.msg_controllen = 0;
+ msgvec->msg_hdr.msg_flags = 0;
+ msgvec++;
+ }
+ return result;
+}
+
+static void net_udst_cleanup(NetClientState *nc)
+{
+ NetUdstState *s = DO_UPCAST(NetUdstState, nc, nc);
+ qemu_purge_queued_packets(nc);
+ udst_read_poll(s, false);
+ udst_write_poll(s, false);
+ if (s->fd >= 0) {
+ close(s->fd);
+ }
+ if (s->header_size > 0) {
+ destroy_vector(s->msgvec, MAX_UNIFIED_MSGCNT, IOVSIZE);
+ } else {
+ destroy_vector(s->msgvec, MAX_UNIFIED_MSGCNT, 1);
+ }
+ g_free(s->vec);
+ if (s->header_buf != NULL) {
+ g_free(s->header_buf);
+ }
+ if (s->dgram_dst != NULL) {
+ g_free(s->dgram_dst);
+ }
+}
+
+static NetClientInfo net_udst_info = {
+ /* we share this one for all types for now, wrong I know :) */
+ .type = NET_CLIENT_DRIVER_UDST,
+ .size = sizeof(NetUdstState),
+ .receive = net_udst_receive_dgram,
+ .receive_iov = net_udst_receive_dgram_iov,
+ .poll = udst_poll,
+ .cleanup = net_udst_cleanup,
+};
+
+NetClientState *qemu_new_udst_net_client(const char *name,
+ NetClientState *peer) {
+ return qemu_new_net_client(&net_udst_info, peer, "udst", name);
+}
+
+void qemu_net_finalize_udst_init(NetUdstState *s,
+ int (*verify_header)(void *s, uint8_t *buf),
+ void (*form_header)(void *s),
+ int fd)
+{
+
+ s->form_header = form_header;
+ s->verify_header = verify_header;
+ s->queue_head = 0;
+ s->queue_tail = 0;
+ s->header_mismatch = false;
+ s->msgvec = build_udst_vector(s, MAX_UNIFIED_MSGCNT);
+ s->vec = g_new(struct iovec, MAX_UNIFIED_IOVCNT);
+ if (s->header_size > 0) {
+ s->header_buf = g_malloc(s->header_size);
+ } else {
+ s->header_buf = NULL;
+ }
+ qemu_set_nonblock(fd);
+
+ s->fd = fd;
+ udst_read_poll(s, true);
+
+}
diff --git a/net/udst.h b/net/udst.h
new file mode 100644
index 0000000000..2a6b44c74d
--- /dev/null
+++ b/net/udst.h
@@ -0,0 +1,121 @@
+/*
+ * QEMU System Emulator
+ *
+ * Copyright (c) 2015-2017 Cambridge Greys Limited
+ * Copyright (c) 2012-2014 Cisco Systems
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+
+
+#define BUFFER_ALIGN sysconf(_SC_PAGESIZE)
+#define BUFFER_SIZE 2048
+#define IOVSIZE 2
+#define MAX_UNIFIED_MSGCNT 64
+#define MAX_UNIFIED_IOVCNT (MAX_UNIFIED_MSGCNT * IOVSIZE)
+
+#ifndef QEMU_NET_UNIFIED_H
+#define QEMU_NET_UNIFIED_H
+
+typedef struct NetUdstState {
+ NetClientState nc;
+
+ int fd;
+
+ /*
+ * these are used for xmit - that happens packet a time
+ * and for first sign of life packet (easier to parse that once)
+ */
+
+ uint8_t *header_buf;
+ struct iovec *vec;
+
+ /*
+ * these are used for receive - try to "eat" up to 32 packets at a time
+ */
+
+ struct mmsghdr *msgvec;
+
+ /*
+ * peer address
+ */
+
+ struct sockaddr_storage *dgram_dst;
+ uint32_t dst_size;
+
+ /*
+ * Internal Queue
+ */
+
+ /*
+ * DOS avoidance in error handling
+ */
+
+ /* Easier to keep l2tpv3 specific */
+
+ bool header_mismatch;
+
+ /*
+ *
+ * Ring buffer handling
+ *
+ */
+
+ int queue_head;
+ int queue_tail;
+ int queue_depth;
+
+ /*
+ * Offset to data - common for all protocols
+ */
+
+ uint32_t offset;
+
+ /*
+ * Header size - common for all protocols
+ */
+
+ uint32_t header_size;
+ /* Poll Control */
+
+ bool read_poll;
+ bool write_poll;
+
+ /* Parameters */
+
+ void *params;
+
+ /* header forming functions */
+
+ int (*verify_header)(void *s, uint8_t *buf);
+ void (*form_header)(void *s);
+
+} NetUdstState;
+
+extern NetClientState *qemu_new_udst_net_client(const char *name,
+ NetClientState *peer);
+
+extern void qemu_net_finalize_udst_init(NetUdstState *s,
+ int (*verify_header)(void *s, uint8_t *buf),
+ void (*form_header)(void *s),
+ int fd);
+#endif
diff --git a/qapi-schema.json b/qapi-schema.json
index 8b015bee2e..62a044f006 100644
--- a/qapi-schema.json
+++ b/qapi-schema.json
@@ -3854,6 +3854,18 @@
'*offset': 'uint32' } }
##
+# @NetdevUdstOptions:
+#
+# Common Datagram backend for unified datagram
+# socket transports
+# Should not be instantiated directly
+
+# Since: 2.11
+##
+{ 'struct': 'NetdevUdstOptions',
+ 'data': { } }
+
+##
# @NetdevVdeOptions:
#
# Connect the VLAN to a vde switch running on the host.
@@ -3971,7 +3983,7 @@
##
{ 'enum': 'NetClientDriver',
'data': [ 'none', 'nic', 'user', 'tap', 'l2tpv3', 'socket', 'vde', 'dump',
- 'bridge', 'hubport', 'netmap', 'vhost-user' ] }
+ 'bridge', 'hubport', 'netmap', 'vhost-user', 'udst' ] }
##
# @Netdev:
@@ -3985,6 +3997,8 @@
# Since: 1.2
#
# 'l2tpv3' - since 2.1
+#
+# 'udst' - since 2.11
##
{ 'union': 'Netdev',
'base': { 'id': 'str', 'type': 'NetClientDriver' },
@@ -4001,7 +4015,8 @@
'bridge': 'NetdevBridgeOptions',
'hubport': 'NetdevHubPortOptions',
'netmap': 'NetdevNetmapOptions',
- 'vhost-user': 'NetdevVhostUserOptions' } }
+ 'vhost-user': 'NetdevVhostUserOptions',
+ 'udst': 'NetdevUdstOptions' } }
##
# @NetLegacy:
diff --git a/qemu-options.hx b/qemu-options.hx
index 746b5fa75d..9caf53fd76 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -1945,7 +1945,7 @@ DEF("netdev", HAS_ARG, QEMU_OPTION_netdev,
" connected to a bridge (default=" DEFAULT_BRIDGE_INTERFACE ")\n"
" using the program 'helper (default=" DEFAULT_BRIDGE_HELPER ")\n"
#endif
-#ifdef __linux__
+#ifdef CONFIG_UDST
"-netdev l2tpv3,id=str,src=srcaddr,dst=dstaddr[,srcport=srcport][,dstport=dstport]\n"
" [,rxsession=rxsession],txsession=txsession[,ipv6=on/off][,udp=on/off]\n"
" [,cookie64=on/off][,counter][,pincounter][,txcookie=txcookie]\n"
--
2.11.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [Qemu-devel] [PATCH 2/4] Migrate l2tpv3 to UDST Backend
2017-07-19 20:02 [Qemu-devel] Revised Unified Datagram Socket Transport patchset anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 1/4] Unified Datagram Socket Transports anton.ivanov
@ 2017-07-19 20:02 ` anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 3/4] GRETAP Backend for UDST anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 4/4] Raw " anton.ivanov
3 siblings, 0 replies; 5+ messages in thread
From: anton.ivanov @ 2017-07-19 20:02 UTC (permalink / raw)
To: qemu-devel; +Cc: jasowang, Anton Ivanov
From: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Migrate L2TPv3 transport to the Unified Datagram Socket
Transport Backend.
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
---
net/l2tpv3.c | 537 +++++++++--------------------------------------------------
1 file changed, 83 insertions(+), 454 deletions(-)
diff --git a/net/l2tpv3.c b/net/l2tpv3.c
index 6745b78990..25b7628244 100644
--- a/net/l2tpv3.c
+++ b/net/l2tpv3.c
@@ -1,6 +1,7 @@
/*
* QEMU System Emulator
*
+ * Copyright (c) 2015-2017 Cambridge Greys Limited
* Copyright (c) 2003-2008 Fabrice Bellard
* Copyright (c) 2012-2014 Cisco Systems
*
@@ -30,23 +31,14 @@
#include "clients.h"
#include "qemu-common.h"
#include "qemu/error-report.h"
+#include "qapi/error.h"
#include "qemu/option.h"
#include "qemu/sockets.h"
#include "qemu/iov.h"
#include "qemu/main-loop.h"
+#include "udst.h"
-/* The buffer size needs to be investigated for optimum numbers and
- * optimum means of paging in on different systems. This size is
- * chosen to be sufficient to accommodate one packet with some headers
- */
-
-#define BUFFER_ALIGN sysconf(_SC_PAGESIZE)
-#define BUFFER_SIZE 2048
-#define IOVSIZE 2
-#define MAX_L2TPV3_MSGCNT 64
-#define MAX_L2TPV3_IOVCNT (MAX_L2TPV3_MSGCNT * IOVSIZE)
-
/* Header set to 0x30000 signifies a data packet */
#define L2TPV3_DATA_PACKET 0x30000
@@ -57,31 +49,7 @@
#define IPPROTO_L2TP 0x73
#endif
-typedef struct NetL2TPV3State {
- NetClientState nc;
- int fd;
-
- /*
- * these are used for xmit - that happens packet a time
- * and for first sign of life packet (easier to parse that once)
- */
-
- uint8_t *header_buf;
- struct iovec *vec;
-
- /*
- * these are used for receive - try to "eat" up to 32 packets at a time
- */
-
- struct mmsghdr *msgvec;
-
- /*
- * peer address
- */
-
- struct sockaddr_storage *dgram_dst;
- uint32_t dst_size;
-
+typedef struct L2TPV3TunnelParams {
/*
* L2TPv3 parameters
*/
@@ -90,37 +58,8 @@ typedef struct NetL2TPV3State {
uint64_t tx_cookie;
uint32_t rx_session;
uint32_t tx_session;
- uint32_t header_size;
uint32_t counter;
- /*
- * DOS avoidance in error handling
- */
-
- bool header_mismatch;
-
- /*
- * Ring buffer handling
- */
-
- int queue_head;
- int queue_tail;
- int queue_depth;
-
- /*
- * Precomputed offsets
- */
-
- uint32_t offset;
- uint32_t cookie_offset;
- uint32_t counter_offset;
- uint32_t session_offset;
-
- /* Poll Control */
-
- bool read_poll;
- bool write_poll;
-
/* Flags */
bool ipv6;
@@ -130,189 +69,62 @@ typedef struct NetL2TPV3State {
bool cookie;
bool cookie_is_64;
-} NetL2TPV3State;
-
-static void net_l2tpv3_send(void *opaque);
-static void l2tpv3_writable(void *opaque);
-
-static void l2tpv3_update_fd_handler(NetL2TPV3State *s)
-{
- qemu_set_fd_handler(s->fd,
- s->read_poll ? net_l2tpv3_send : NULL,
- s->write_poll ? l2tpv3_writable : NULL,
- s);
-}
-
-static void l2tpv3_read_poll(NetL2TPV3State *s, bool enable)
-{
- if (s->read_poll != enable) {
- s->read_poll = enable;
- l2tpv3_update_fd_handler(s);
- }
-}
+ /* Precomputed L2TPV3 specific offsets */
+ uint32_t cookie_offset;
+ uint32_t counter_offset;
+ uint32_t session_offset;
-static void l2tpv3_write_poll(NetL2TPV3State *s, bool enable)
-{
- if (s->write_poll != enable) {
- s->write_poll = enable;
- l2tpv3_update_fd_handler(s);
- }
-}
+} L2TPV3TunnelParams;
-static void l2tpv3_writable(void *opaque)
-{
- NetL2TPV3State *s = opaque;
- l2tpv3_write_poll(s, false);
- qemu_flush_queued_packets(&s->nc);
-}
-static void l2tpv3_send_completed(NetClientState *nc, ssize_t len)
-{
- NetL2TPV3State *s = DO_UPCAST(NetL2TPV3State, nc, nc);
- l2tpv3_read_poll(s, true);
-}
-static void l2tpv3_poll(NetClientState *nc, bool enable)
+static void l2tpv3_form_header(void *us)
{
- NetL2TPV3State *s = DO_UPCAST(NetL2TPV3State, nc, nc);
- l2tpv3_write_poll(s, enable);
- l2tpv3_read_poll(s, enable);
-}
+ NetUdstState *s = (NetUdstState *) us;
+ L2TPV3TunnelParams *p = (L2TPV3TunnelParams *) s->params;
-static void l2tpv3_form_header(NetL2TPV3State *s)
-{
uint32_t *counter;
- if (s->udp) {
+ if (p->udp) {
stl_be_p((uint32_t *) s->header_buf, L2TPV3_DATA_PACKET);
}
stl_be_p(
- (uint32_t *) (s->header_buf + s->session_offset),
- s->tx_session
+ (uint32_t *) (s->header_buf + p->session_offset),
+ p->tx_session
);
- if (s->cookie) {
- if (s->cookie_is_64) {
+ if (p->cookie) {
+ if (p->cookie_is_64) {
stq_be_p(
- (uint64_t *)(s->header_buf + s->cookie_offset),
- s->tx_cookie
+ (uint64_t *)(s->header_buf + p->cookie_offset),
+ p->tx_cookie
);
} else {
stl_be_p(
- (uint32_t *) (s->header_buf + s->cookie_offset),
- s->tx_cookie
+ (uint32_t *) (s->header_buf + p->cookie_offset),
+ p->tx_cookie
);
}
}
- if (s->has_counter) {
- counter = (uint32_t *)(s->header_buf + s->counter_offset);
- if (s->pin_counter) {
+ if (p->has_counter) {
+ counter = (uint32_t *)(s->header_buf + p->counter_offset);
+ if (p->pin_counter) {
*counter = 0;
} else {
- stl_be_p(counter, ++s->counter);
+ stl_be_p(counter, ++p->counter);
}
}
}
-static ssize_t net_l2tpv3_receive_dgram_iov(NetClientState *nc,
- const struct iovec *iov,
- int iovcnt)
-{
- NetL2TPV3State *s = DO_UPCAST(NetL2TPV3State, nc, nc);
- struct msghdr message;
- int ret;
-
- if (iovcnt > MAX_L2TPV3_IOVCNT - 1) {
- error_report(
- "iovec too long %d > %d, change l2tpv3.h",
- iovcnt, MAX_L2TPV3_IOVCNT
- );
- return -1;
- }
- l2tpv3_form_header(s);
- memcpy(s->vec + 1, iov, iovcnt * sizeof(struct iovec));
- s->vec->iov_base = s->header_buf;
- s->vec->iov_len = s->offset;
- message.msg_name = s->dgram_dst;
- message.msg_namelen = s->dst_size;
- message.msg_iov = s->vec;
- message.msg_iovlen = iovcnt + 1;
- message.msg_control = NULL;
- message.msg_controllen = 0;
- message.msg_flags = 0;
- do {
- ret = sendmsg(s->fd, &message, 0);
- } while ((ret == -1) && (errno == EINTR));
- if (ret > 0) {
- ret -= s->offset;
- } else if (ret == 0) {
- /* belt and braces - should not occur on DGRAM
- * we should get an error and never a 0 send
- */
- ret = iov_size(iov, iovcnt);
- } else {
- /* signal upper layer that socket buffer is full */
- ret = -errno;
- if (ret == -EAGAIN || ret == -ENOBUFS) {
- l2tpv3_write_poll(s, true);
- ret = 0;
- }
- }
- return ret;
-}
-
-static ssize_t net_l2tpv3_receive_dgram(NetClientState *nc,
- const uint8_t *buf,
- size_t size)
-{
- NetL2TPV3State *s = DO_UPCAST(NetL2TPV3State, nc, nc);
-
- struct iovec *vec;
- struct msghdr message;
- ssize_t ret = 0;
-
- l2tpv3_form_header(s);
- vec = s->vec;
- vec->iov_base = s->header_buf;
- vec->iov_len = s->offset;
- vec++;
- vec->iov_base = (void *) buf;
- vec->iov_len = size;
- message.msg_name = s->dgram_dst;
- message.msg_namelen = s->dst_size;
- message.msg_iov = s->vec;
- message.msg_iovlen = 2;
- message.msg_control = NULL;
- message.msg_controllen = 0;
- message.msg_flags = 0;
- do {
- ret = sendmsg(s->fd, &message, 0);
- } while ((ret == -1) && (errno == EINTR));
- if (ret > 0) {
- ret -= s->offset;
- } else if (ret == 0) {
- /* belt and braces - should not occur on DGRAM
- * we should get an error and never a 0 send
- */
- ret = size;
- } else {
- ret = -errno;
- if (ret == -EAGAIN || ret == -ENOBUFS) {
- /* signal upper layer that socket buffer is full */
- l2tpv3_write_poll(s, true);
- ret = 0;
- }
- }
- return ret;
-}
-
-static int l2tpv3_verify_header(NetL2TPV3State *s, uint8_t *buf)
+static int l2tpv3_verify_header(void *us, uint8_t *buf)
{
+ NetUdstState *s = (NetUdstState *) us;
+ L2TPV3TunnelParams *p = (L2TPV3TunnelParams *) s->params;
uint32_t *session;
uint64_t cookie;
- if ((!s->udp) && (!s->ipv6)) {
+ if ((!p->udp) && (!p->ipv6)) {
buf += sizeof(struct iphdr) /* fix for ipv4 raw */;
}
@@ -321,21 +133,21 @@ static int l2tpv3_verify_header(NetL2TPV3State *s, uint8_t *buf)
* that anyway.
*/
- if (s->cookie) {
- if (s->cookie_is_64) {
- cookie = ldq_be_p(buf + s->cookie_offset);
+ if (p->cookie) {
+ if (p->cookie_is_64) {
+ cookie = ldq_be_p(buf + p->cookie_offset);
} else {
- cookie = ldl_be_p(buf + s->cookie_offset) & 0xffffffffULL;
+ cookie = ldl_be_p(buf + p->cookie_offset) & 0xffffffffULL;
}
- if (cookie != s->rx_cookie) {
+ if (cookie != p->rx_cookie) {
if (!s->header_mismatch) {
error_report("unknown cookie id");
}
return -1;
}
}
- session = (uint32_t *) (buf + s->session_offset);
- if (ldl_be_p(session) != s->rx_session) {
+ session = (uint32_t *) (buf + p->session_offset);
+ if (ldl_be_p(session) != p->rx_session) {
if (!s->header_mismatch) {
error_report("session mismatch");
}
@@ -344,214 +156,35 @@ static int l2tpv3_verify_header(NetL2TPV3State *s, uint8_t *buf)
return 0;
}
-static void net_l2tpv3_process_queue(NetL2TPV3State *s)
-{
- int size = 0;
- struct iovec *vec;
- bool bad_read;
- int data_size;
- struct mmsghdr *msgvec;
-
- /* go into ring mode only if there is a "pending" tail */
- if (s->queue_depth > 0) {
- do {
- msgvec = s->msgvec + s->queue_tail;
- if (msgvec->msg_len > 0) {
- data_size = msgvec->msg_len - s->header_size;
- vec = msgvec->msg_hdr.msg_iov;
- if ((data_size > 0) &&
- (l2tpv3_verify_header(s, vec->iov_base) == 0)) {
- vec++;
- /* Use the legacy delivery for now, we will
- * switch to using our own ring as a queueing mechanism
- * at a later date
- */
- size = qemu_send_packet_async(
- &s->nc,
- vec->iov_base,
- data_size,
- l2tpv3_send_completed
- );
- if (size == 0) {
- l2tpv3_read_poll(s, false);
- }
- bad_read = false;
- } else {
- bad_read = true;
- if (!s->header_mismatch) {
- /* report error only once */
- error_report("l2tpv3 header verification failed");
- s->header_mismatch = true;
- }
- }
- } else {
- bad_read = true;
- }
- s->queue_tail = (s->queue_tail + 1) % MAX_L2TPV3_MSGCNT;
- s->queue_depth--;
- } while (
- (s->queue_depth > 0) &&
- qemu_can_send_packet(&s->nc) &&
- ((size > 0) || bad_read)
- );
- }
-}
-
-static void net_l2tpv3_send(void *opaque)
-{
- NetL2TPV3State *s = opaque;
- int target_count, count;
- struct mmsghdr *msgvec;
-
- /* go into ring mode only if there is a "pending" tail */
-
- if (s->queue_depth) {
-
- /* The ring buffer we use has variable intake
- * count of how much we can read varies - adjust accordingly
- */
-
- target_count = MAX_L2TPV3_MSGCNT - s->queue_depth;
-
- /* Ensure we do not overrun the ring when we have
- * a lot of enqueued packets
- */
-
- if (s->queue_head + target_count > MAX_L2TPV3_MSGCNT) {
- target_count = MAX_L2TPV3_MSGCNT - s->queue_head;
- }
- } else {
-
- /* we do not have any pending packets - we can use
- * the whole message vector linearly instead of using
- * it as a ring
- */
-
- s->queue_head = 0;
- s->queue_tail = 0;
- target_count = MAX_L2TPV3_MSGCNT;
- }
-
- msgvec = s->msgvec + s->queue_head;
- if (target_count > 0) {
- do {
- count = recvmmsg(
- s->fd,
- msgvec,
- target_count, MSG_DONTWAIT, NULL);
- } while ((count == -1) && (errno == EINTR));
- if (count < 0) {
- /* Recv error - we still need to flush packets here,
- * (re)set queue head to current position
- */
- count = 0;
- }
- s->queue_head = (s->queue_head + count) % MAX_L2TPV3_MSGCNT;
- s->queue_depth += count;
- }
- net_l2tpv3_process_queue(s);
-}
-
-static void destroy_vector(struct mmsghdr *msgvec, int count, int iovcount)
-{
- int i, j;
- struct iovec *iov;
- struct mmsghdr *cleanup = msgvec;
- if (cleanup) {
- for (i = 0; i < count; i++) {
- if (cleanup->msg_hdr.msg_iov) {
- iov = cleanup->msg_hdr.msg_iov;
- for (j = 0; j < iovcount; j++) {
- g_free(iov->iov_base);
- iov++;
- }
- g_free(cleanup->msg_hdr.msg_iov);
- }
- cleanup++;
- }
- g_free(msgvec);
- }
-}
-
-static struct mmsghdr *build_l2tpv3_vector(NetL2TPV3State *s, int count)
-{
- int i;
- struct iovec *iov;
- struct mmsghdr *msgvec, *result;
-
- msgvec = g_new(struct mmsghdr, count);
- result = msgvec;
- for (i = 0; i < count ; i++) {
- msgvec->msg_hdr.msg_name = NULL;
- msgvec->msg_hdr.msg_namelen = 0;
- iov = g_new(struct iovec, IOVSIZE);
- msgvec->msg_hdr.msg_iov = iov;
- iov->iov_base = g_malloc(s->header_size);
- iov->iov_len = s->header_size;
- iov++ ;
- iov->iov_base = qemu_memalign(BUFFER_ALIGN, BUFFER_SIZE);
- iov->iov_len = BUFFER_SIZE;
- msgvec->msg_hdr.msg_iovlen = 2;
- msgvec->msg_hdr.msg_control = NULL;
- msgvec->msg_hdr.msg_controllen = 0;
- msgvec->msg_hdr.msg_flags = 0;
- msgvec++;
- }
- return result;
-}
-
-static void net_l2tpv3_cleanup(NetClientState *nc)
-{
- NetL2TPV3State *s = DO_UPCAST(NetL2TPV3State, nc, nc);
- qemu_purge_queued_packets(nc);
- l2tpv3_read_poll(s, false);
- l2tpv3_write_poll(s, false);
- if (s->fd >= 0) {
- close(s->fd);
- }
- destroy_vector(s->msgvec, MAX_L2TPV3_MSGCNT, IOVSIZE);
- g_free(s->vec);
- g_free(s->header_buf);
- g_free(s->dgram_dst);
-}
-
-static NetClientInfo net_l2tpv3_info = {
- .type = NET_CLIENT_DRIVER_L2TPV3,
- .size = sizeof(NetL2TPV3State),
- .receive = net_l2tpv3_receive_dgram,
- .receive_iov = net_l2tpv3_receive_dgram_iov,
- .poll = l2tpv3_poll,
- .cleanup = net_l2tpv3_cleanup,
-};
-
int net_init_l2tpv3(const Netdev *netdev,
const char *name,
NetClientState *peer, Error **errp)
{
- /* FIXME error_setg(errp, ...) on failure */
const NetdevL2TPv3Options *l2tpv3;
- NetL2TPV3State *s;
+ NetUdstState *s;
NetClientState *nc;
+ L2TPV3TunnelParams *p;
+
int fd = -1, gairet;
struct addrinfo hints;
struct addrinfo *result = NULL;
char *srcport, *dstport;
- nc = qemu_new_net_client(&net_l2tpv3_info, peer, "l2tpv3", name);
+ nc = qemu_new_udst_net_client(name, peer);
+
+ s = DO_UPCAST(NetUdstState, nc, nc);
- s = DO_UPCAST(NetL2TPV3State, nc, nc);
+ p = g_malloc(sizeof(L2TPV3TunnelParams));
- s->queue_head = 0;
- s->queue_tail = 0;
- s->header_mismatch = false;
+ s->params = p;
assert(netdev->type == NET_CLIENT_DRIVER_L2TPV3);
l2tpv3 = &netdev->u.l2tpv3;
if (l2tpv3->has_ipv6 && l2tpv3->ipv6) {
- s->ipv6 = l2tpv3->ipv6;
+ p->ipv6 = l2tpv3->ipv6;
} else {
- s->ipv6 = false;
+ p->ipv6 = false;
}
if ((l2tpv3->has_offset) && (l2tpv3->offset > 256)) {
@@ -561,22 +194,22 @@ int net_init_l2tpv3(const Netdev *netdev,
if (l2tpv3->has_rxcookie || l2tpv3->has_txcookie) {
if (l2tpv3->has_rxcookie && l2tpv3->has_txcookie) {
- s->cookie = true;
+ p->cookie = true;
} else {
goto outerr;
}
} else {
- s->cookie = false;
+ p->cookie = false;
}
if (l2tpv3->has_cookie64 || l2tpv3->cookie64) {
- s->cookie_is_64 = true;
+ p->cookie_is_64 = true;
} else {
- s->cookie_is_64 = false;
+ p->cookie_is_64 = false;
}
if (l2tpv3->has_udp && l2tpv3->udp) {
- s->udp = true;
+ p->udp = true;
if (!(l2tpv3->has_srcport && l2tpv3->has_dstport)) {
error_report("l2tpv3_open : need both src and dst port for udp");
goto outerr;
@@ -585,52 +218,52 @@ int net_init_l2tpv3(const Netdev *netdev,
dstport = l2tpv3->dstport;
}
} else {
- s->udp = false;
+ p->udp = false;
srcport = NULL;
dstport = NULL;
}
s->offset = 4;
- s->session_offset = 0;
- s->cookie_offset = 4;
- s->counter_offset = 4;
+ p->session_offset = 0;
+ p->cookie_offset = 4;
+ p->counter_offset = 4;
- s->tx_session = l2tpv3->txsession;
+ p->tx_session = l2tpv3->txsession;
if (l2tpv3->has_rxsession) {
- s->rx_session = l2tpv3->rxsession;
+ p->rx_session = l2tpv3->rxsession;
} else {
- s->rx_session = s->tx_session;
+ p->rx_session = p->tx_session;
}
- if (s->cookie) {
- s->rx_cookie = l2tpv3->rxcookie;
- s->tx_cookie = l2tpv3->txcookie;
- if (s->cookie_is_64 == true) {
+ if (p->cookie) {
+ p->rx_cookie = l2tpv3->rxcookie;
+ p->tx_cookie = l2tpv3->txcookie;
+ if (p->cookie_is_64 == true) {
/* 64 bit cookie */
s->offset += 8;
- s->counter_offset += 8;
+ p->counter_offset += 8;
} else {
/* 32 bit cookie */
s->offset += 4;
- s->counter_offset += 4;
+ p->counter_offset += 4;
}
}
memset(&hints, 0, sizeof(hints));
- if (s->ipv6) {
+ if (p->ipv6) {
hints.ai_family = AF_INET6;
} else {
hints.ai_family = AF_INET;
}
- if (s->udp) {
+ if (p->udp) {
hints.ai_socktype = SOCK_DGRAM;
hints.ai_protocol = 0;
s->offset += 4;
- s->counter_offset += 4;
- s->session_offset += 4;
- s->cookie_offset += 4;
+ p->counter_offset += 4;
+ p->session_offset += 4;
+ p->cookie_offset += 4;
} else {
hints.ai_socktype = SOCK_RAW;
hints.ai_protocol = IPPROTO_L2TP;
@@ -661,12 +294,12 @@ int net_init_l2tpv3(const Netdev *netdev,
memset(&hints, 0, sizeof(hints));
- if (s->ipv6) {
+ if (p->ipv6) {
hints.ai_family = AF_INET6;
} else {
hints.ai_family = AF_INET;
}
- if (s->udp) {
+ if (p->udp) {
hints.ai_socktype = SOCK_DGRAM;
hints.ai_protocol = 0;
} else {
@@ -693,17 +326,17 @@ int net_init_l2tpv3(const Netdev *netdev,
}
if (l2tpv3->has_counter && l2tpv3->counter) {
- s->has_counter = true;
+ p->has_counter = true;
s->offset += 4;
} else {
- s->has_counter = false;
+ p->has_counter = false;
}
if (l2tpv3->has_pincounter && l2tpv3->pincounter) {
- s->has_counter = true; /* pin counter implies that there is counter */
- s->pin_counter = true;
+ p->has_counter = true; /* pin counter implies that there is counter */
+ p->pin_counter = true;
} else {
- s->pin_counter = false;
+ p->pin_counter = false;
}
if (l2tpv3->has_offset) {
@@ -711,27 +344,23 @@ int net_init_l2tpv3(const Netdev *netdev,
s->offset += l2tpv3->offset;
}
- if ((s->ipv6) || (s->udp)) {
+ if ((p->ipv6) || (p->udp)) {
s->header_size = s->offset;
} else {
s->header_size = s->offset + sizeof(struct iphdr);
}
- s->msgvec = build_l2tpv3_vector(s, MAX_L2TPV3_MSGCNT);
- s->vec = g_new(struct iovec, MAX_L2TPV3_IOVCNT);
- s->header_buf = g_malloc(s->header_size);
-
- qemu_set_nonblock(fd);
-
- s->fd = fd;
- s->counter = 0;
-
- l2tpv3_read_poll(s, true);
+ qemu_net_finalize_udst_init(s,
+ &l2tpv3_verify_header,
+ &l2tpv3_form_header,
+ fd);
+ p->counter = 0;
snprintf(s->nc.info_str, sizeof(s->nc.info_str),
"l2tpv3: connected");
return 0;
outerr:
+ error_setg(errp, "Cannot initialize L2TPv3 transport");
qemu_del_net_client(nc);
if (fd >= 0) {
close(fd);
--
2.11.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [Qemu-devel] [PATCH 3/4] GRETAP Backend for UDST
2017-07-19 20:02 [Qemu-devel] Revised Unified Datagram Socket Transport patchset anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 1/4] Unified Datagram Socket Transports anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 2/4] Migrate l2tpv3 to UDST Backend anton.ivanov
@ 2017-07-19 20:02 ` anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 4/4] Raw " anton.ivanov
3 siblings, 0 replies; 5+ messages in thread
From: anton.ivanov @ 2017-07-19 20:02 UTC (permalink / raw)
To: qemu-devel; +Cc: jasowang, Anton Ivanov
From: Anton Ivanov <anton.ivanov@cambridgegreys.com>
GRETAP Backend for Universal Datagram Socket Transport
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
---
net/Makefile.objs | 2 +-
net/clients.h | 4 +
net/gre.c | 311 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
net/net.c | 1 +
qapi-schema.json | 41 ++++++-
qemu-options.hx | 60 ++++++++++-
6 files changed, 414 insertions(+), 5 deletions(-)
create mode 100644 net/gre.c
diff --git a/net/Makefile.objs b/net/Makefile.objs
index ffdfb96bd0..919bc3d78f 100644
--- a/net/Makefile.objs
+++ b/net/Makefile.objs
@@ -2,7 +2,7 @@ common-obj-y = net.o queue.o checksum.o util.o hub.o
common-obj-y += socket.o
common-obj-y += dump.o
common-obj-y += eth.o
-common-obj-$(CONFIG_UDST) += udst.o l2tpv3.o
+common-obj-$(CONFIG_UDST) += udst.o l2tpv3.o gre.o
common-obj-$(CONFIG_POSIX) += vhost-user.o
common-obj-$(CONFIG_SLIRP) += slirp.o
common-obj-$(CONFIG_VDE) += vde.o
diff --git a/net/clients.h b/net/clients.h
index 5cae479730..8f8a59aee3 100644
--- a/net/clients.h
+++ b/net/clients.h
@@ -49,6 +49,10 @@ int net_init_bridge(const Netdev *netdev, const char *name,
int net_init_l2tpv3(const Netdev *netdev, const char *name,
NetClientState *peer, Error **errp);
+
+int net_init_gre(const Netdev *netdev, const char *name,
+ NetClientState *peer, Error **errp);
+
#ifdef CONFIG_VDE
int net_init_vde(const Netdev *netdev, const char *name,
NetClientState *peer, Error **errp);
diff --git a/net/gre.c b/net/gre.c
new file mode 100644
index 0000000000..7734d78102
--- /dev/null
+++ b/net/gre.c
@@ -0,0 +1,311 @@
+/*
+ * QEMU System Emulator
+ *
+ * Copyright (c) 2015-2017 Cambridge GREys Limited
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ * Copyright (c) 2012-2014 Cisco Systems
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include <linux/ip.h>
+#include <netdb.h>
+#include "net/net.h"
+#include "clients.h"
+#include "qemu-common.h"
+#include "qemu/error-report.h"
+#include "qapi/error.h"
+#include "qemu/option.h"
+#include "qemu/sockets.h"
+#include "qemu/iov.h"
+#include "qemu/main-loop.h"
+#include "udst.h"
+
+/* IANA-assigned IP protocol ID for GRE */
+
+
+#ifndef IPPROTO_GRE
+#define IPPROTO_GRE 0x2F
+#endif
+
+#define GRE_MODE_CHECKSUM htons(8 << 12) /* checksum */
+#define GRE_MODE_RESERVED htons(4 << 12) /* unused */
+#define GRE_MODE_KEY htons(2 << 12) /* KEY present */
+#define GRE_MODE_SEQUENCE htons(1 << 12) /* no sequence */
+
+
+/* GRE TYPE for Ethernet in GRE aka GRETAP */
+
+#define GRE_IRB htons(0x6558)
+
+struct gre_minimal_header {
+ uint16_t header;
+ uint16_t arptype;
+};
+
+typedef struct GRETunnelParams {
+ /*
+ * GRE parameters
+ */
+
+ uint32_t rx_key;
+ uint32_t tx_key;
+ uint32_t sequence;
+
+ /* Flags */
+
+ bool ipv6;
+ bool udp;
+ bool has_sequence;
+ bool pin_sequence;
+ bool checksum;
+ bool key;
+
+ /* Precomputed GRE specific offsets */
+
+ uint32_t key_offset;
+ uint32_t sequence_offset;
+ uint32_t checksum_offset;
+
+ struct gre_minimal_header header_bits;
+
+} GRETunnelParams;
+
+
+
+static void gre_form_header(void *us)
+{
+ NetUdstState *s = (NetUdstState *) us;
+ GRETunnelParams *p = (GRETunnelParams *) s->params;
+
+ uint32_t *sequence;
+
+ *((uint32_t *) s->header_buf) = *((uint32_t *) &p->header_bits);
+
+ if (p->key) {
+ stl_be_p(
+ (uint32_t *) (s->header_buf + p->key_offset),
+ p->tx_key
+ );
+ }
+ if (p->has_sequence) {
+ sequence = (uint32_t *)(s->header_buf + p->sequence_offset);
+ if (p->pin_sequence) {
+ *sequence = 0;
+ } else {
+ stl_be_p(sequence, ++p->sequence);
+ }
+ }
+}
+
+static int gre_verify_header(void *us, uint8_t *buf)
+{
+
+ NetUdstState *s = (NetUdstState *) us;
+ GRETunnelParams *p = (GRETunnelParams *) s->params;
+ uint32_t key;
+
+
+ if (!p->ipv6) {
+ buf += sizeof(struct iphdr) /* fix for ipv4 raw */;
+ }
+
+ if (*((uint32_t *) buf) != *((uint32_t *) &p->header_bits)) {
+ if (!s->header_mismatch) {
+ error_report("header type disagreement, expecting %0x, got %0x",
+ *((uint32_t *) &p->header_bits), *((uint32_t *) buf));
+ }
+ return -1;
+ }
+
+ if (p->key) {
+ key = ldl_be_p(buf + p->key_offset);
+ if (key != p->rx_key) {
+ if (!s->header_mismatch) {
+ error_report("unknown key id %0x, expecting %0x",
+ key, p->rx_key);
+ }
+ return -1;
+ }
+ }
+ return 0;
+}
+
+int net_init_gre(const Netdev *netdev,
+ const char *name,
+ NetClientState *peer, Error **errp)
+{
+ const NetdevGREOptions *gre;
+ NetUdstState *s;
+ NetClientState *nc;
+ GRETunnelParams *p;
+
+ int fd = -1, gairet;
+ struct addrinfo hints;
+ struct addrinfo *result = NULL;
+
+ nc = qemu_new_udst_net_client(name, peer);
+
+ s = DO_UPCAST(NetUdstState, nc, nc);
+
+ p = g_malloc(sizeof(GRETunnelParams));
+
+ s->params = p;
+ p->header_bits.arptype = GRE_IRB;
+ p->header_bits.header = 0;
+
+ assert(netdev->type == NET_CLIENT_DRIVER_GRE);
+ gre = &netdev->u.gre;
+
+ if (gre->has_ipv6 && gre->ipv6) {
+ p->ipv6 = gre->ipv6;
+ } else {
+ p->ipv6 = false;
+ }
+
+ s->offset = 4;
+ p->key_offset = 4;
+ p->sequence_offset = 4;
+ p->checksum_offset = 4;
+
+ if (gre->has_rxkey || gre->has_txkey) {
+ if (gre->has_rxkey && gre->has_txkey) {
+ p->key = true;
+ p->header_bits.header |= GRE_MODE_KEY;
+ } else {
+ goto outerr;
+ }
+ } else {
+ p->key = false;
+ }
+
+ if (p->key) {
+ p->rx_key = gre->rxkey;
+ p->tx_key = gre->txkey;
+ s->offset += 4;
+ p->sequence_offset += 4;
+ }
+
+
+ if (gre->has_sequence && gre->sequence) {
+ s->offset += 4;
+ p->has_sequence = true;
+ p->header_bits.header |= GRE_MODE_SEQUENCE;
+ } else {
+ p->sequence = false;
+ }
+
+ if (gre->has_pinsequence && gre->pinsequence) {
+ /* pin sequence implies that there is sequence */
+ p->has_sequence = true;
+ p->pin_sequence = true;
+ } else {
+ p->pin_sequence = false;
+ }
+
+ memset(&hints, 0, sizeof(hints));
+
+ if (p->ipv6) {
+ hints.ai_family = AF_INET6;
+ } else {
+ hints.ai_family = AF_INET;
+ }
+
+ hints.ai_socktype = SOCK_RAW;
+ hints.ai_protocol = IPPROTO_GRE;
+
+ gairet = getaddrinfo(gre->src, NULL, &hints, &result);
+
+ if ((gairet != 0) || (result == NULL)) {
+ error_report(
+ "gre_open : could not resolve src, errno = %s",
+ gai_strerror(gairet)
+ );
+ goto outerr;
+ }
+ fd = socket(result->ai_family, result->ai_socktype, result->ai_protocol);
+ if (fd == -1) {
+ fd = -errno;
+ error_report("gre_open : socket creation failed, errno = %d", -fd);
+ goto outerr;
+ }
+ if (bind(fd, (struct sockaddr *) result->ai_addr, result->ai_addrlen)) {
+ error_report("gre_open : could not bind socket err=%i", errno);
+ goto outerr;
+ }
+ if (result) {
+ freeaddrinfo(result);
+ }
+
+ memset(&hints, 0, sizeof(hints));
+
+ if (p->ipv6) {
+ hints.ai_family = AF_INET6;
+ } else {
+ hints.ai_family = AF_INET;
+ }
+ hints.ai_socktype = SOCK_RAW;
+ hints.ai_protocol = IPPROTO_GRE;
+
+ result = NULL;
+ gairet = getaddrinfo(gre->dst, NULL, &hints, &result);
+ if ((gairet != 0) || (result == NULL)) {
+ error_report(
+ "gre_open : could not resolve dst, error = %s",
+ gai_strerror(gairet)
+ );
+ goto outerr;
+ }
+
+ s->dgram_dst = g_new0(struct sockaddr_storage, 1);
+ memcpy(s->dgram_dst, result->ai_addr, result->ai_addrlen);
+ s->dst_size = result->ai_addrlen;
+
+ if (result) {
+ freeaddrinfo(result);
+ }
+
+ if ((p->ipv6) || (p->udp)) {
+ s->header_size = s->offset;
+ } else {
+ s->header_size = s->offset + sizeof(struct iphdr);
+ }
+
+ qemu_net_finalize_udst_init(s,
+ &gre_verify_header,
+ &gre_form_header,
+ fd);
+
+ p->sequence = 0;
+
+ snprintf(s->nc.info_str, sizeof(s->nc.info_str),
+ "gre: connected");
+ return 0;
+outerr:
+ error_setg(errp, "Cannot initialize GRE transport");
+ qemu_del_net_client(nc);
+ if (fd >= 0) {
+ close(fd);
+ }
+ if (result) {
+ freeaddrinfo(result);
+ }
+ return -1;
+}
diff --git a/net/net.c b/net/net.c
index 723a256260..6163a8a3af 100644
--- a/net/net.c
+++ b/net/net.c
@@ -962,6 +962,7 @@ static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
#endif
#ifdef CONFIG_UDST
[NET_CLIENT_DRIVER_L2TPV3] = net_init_l2tpv3,
+ [NET_CLIENT_DRIVER_GRE] = net_init_gre,
#endif
};
diff --git a/qapi-schema.json b/qapi-schema.json
index 62a044f006..082f56645a 100644
--- a/qapi-schema.json
+++ b/qapi-schema.json
@@ -3852,7 +3852,41 @@
'txsession': 'uint32',
'*rxsession': 'uint32',
'*offset': 'uint32' } }
-
+##
+# @NetdevGREOptions:
+#
+# Connect the VLAN to Ethernet over Ethernet over GRE (GRETAP) tunnel
+#
+# @src: source address
+#
+# @dst: destination address
+#
+# @ipv6: force the use of ipv6
+#
+# @sequence: have sequence counter
+#
+# @pinsequence: pin sequence counter to zero -
+# workaround for buggy implementations or
+# networks with packet reorder
+#
+# @txkey: 32 bit transmit key
+#
+# @rxkey: 32 bit receive key
+#
+# Note - gre checksums are not supported at present
+#
+#
+# Since 2.9
+##
+{ 'struct': 'NetdevGREOptions',
+ 'data': {
+ 'src': 'str',
+ 'dst': 'str',
+ '*ipv6': 'bool',
+ '*sequence': 'bool',
+ '*pinsequence': 'bool',
+ '*txkey': 'uint32',
+ '*rxkey': 'uint32' } }
##
# @NetdevUdstOptions:
#
@@ -3983,7 +4017,7 @@
##
{ 'enum': 'NetClientDriver',
'data': [ 'none', 'nic', 'user', 'tap', 'l2tpv3', 'socket', 'vde', 'dump',
- 'bridge', 'hubport', 'netmap', 'vhost-user', 'udst' ] }
+ 'bridge', 'hubport', 'netmap', 'vhost-user', 'udst', 'gre' ] }
##
# @Netdev:
@@ -4016,7 +4050,8 @@
'hubport': 'NetdevHubPortOptions',
'netmap': 'NetdevNetmapOptions',
'vhost-user': 'NetdevVhostUserOptions',
- 'udst': 'NetdevUdstOptions' } }
+ 'udst': 'NetdevUdstOptions',
+ 'gre': 'NetdevGREOptions' } }
##
# @NetLegacy:
diff --git a/qemu-options.hx b/qemu-options.hx
index 9caf53fd76..e35ef032cf 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -1971,6 +1971,23 @@ DEF("netdev", HAS_ARG, QEMU_OPTION_netdev,
" use 'counter=off' to force a 'cut-down' L2TPv3 with no counter\n"
" use 'pincounter=on' to work around broken counter handling in peer\n"
" use 'offset=X' to add an extra offset between header and data\n"
+ "-netdev gre,id=str,src=srcaddr,dst=dstaddr[,rxkey=rxkey],txkey=txkey[,ipv6=on/off]\n"
+ " [,sequence][,pinsequence]\n"
+ " configure a network backend with ID 'str' connected to\n"
+ " an Ethernet over GRE pseudowire (aka GRE TAP).\n"
+ " Linux kernel 3.3+ as well as most routers and some switches\n"
+ " can talk GRETAP. This transport allows connecting a VM to a VM,\n"
+ " VM to a router and even VM to Host. It is a nearly-universal\n"
+ " standard (RFC1701).\n"
+ " use 'src=' to specify source address\n"
+ " use 'dst=' to specify destination address\n"
+ " use 'ipv6=on' to force v6\n"
+ " GRE may use keys to prevent misconfiguration as\n"
+ " well as a weak security measure\n"
+ " use 'rxkey=0x01234' to specify a rxkey\n"
+ " use 'txkey=0x01234' to specify a txkey\n"
+ " use 'sequence=on' to add frame sequence to each packet\n"
+ " use 'pinsequence=on' to work around broken sequence handling in peer\n"
#endif
"-netdev socket,id=str[,fd=h][,listen=[host]:port][,connect=host:port]\n"
" configure a network backend to connect to another network\n"
@@ -2394,12 +2411,53 @@ ip l2tp add session tunnel_id 1 name vmtunnel0 session_id \
ifconfig vmtunnel0 mtu 1500
ifconfig vmtunnel0 up
brctl addif br-lan vmtunnel0
+@end example
+
+Alternatively, it is possible to assign an IP address to vmtunnel0, which allows
+the VM to connect to the host directly without using Linux bridging.
+
+
+@item -netdev gre,id=@var{id},src=@var{srcaddr},dst=@var{dstaddr}[,ipv6][,sequence][,pinsequence][,txkey=@var{txkey}][,rxkey=@var{rxkey}]
+Connect VLAN @var{n} to a GRE pseudowire. GRE (RFC1701) is a popular
+protocol to transport various data frames between two systems.
+We are interested in a specific GRE variety where the transported
+frames are Ethernet. This GRE type is usually referred to as GRETAP.
+It is present in routers, firewalls, switches and the Linux kernel
+(from version 3.3 onwards).
+
+This transport allows a VM to communicate to another VM, router or firewall directly.
+
+@item src=@var{srcaddr}
+ source address (mandatory)
+@item dst=@var{dstaddr}
+ destination address (mandatory)
+@item ipv6
+ force v6, otherwise defaults to v4.
+@item rxkey=@var{rxkey}
+@itemx txkey=@var{txkey}
+ Keys are a weak form of security in the gre specification.
+Their function is mostly to prevent misconfiguration.
+@item sequence=on
+ Add frame sequence to GRE frames
+@item pinsequence=on
+ Work around broken sequence handling in peer. This may also help on
+networks which have packet reorder.
+
+For example, to attach a VM running on host 4.3.2.1 via GRETAP to the bridge br-lan
+on the remote Linux host 1.2.3.4:
+@example
+# Setup tunnel on linux host using raw ip as encapsulation
+# on 1.2.3.4
+ip link add gt0 type gretap local 1.2.3.4 remote 4.3.2.1
+ifconfig gt0 mtu 1500
+ifconfig gt0 up
+brctl addif br-lan gt0
# on 4.3.2.1
# launch QEMU instance - if your network has reorder or is very lossy add ,pincounter
-qemu-system-i386 linux.img -net nic -net l2tpv3,src=4.2.3.1,dst=1.2.3.4,udp,srcport=16384,dstport=16384,rxsession=0xffffffff,txsession=0xffffffff,counter
+qemu-system-i386 linux.img -device virtio-net-pci,netdev=gre0 -netdev gre,id=gre0,src=4.2.3.1,dst=1.2.3.4
@end example
--
2.11.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [Qemu-devel] [PATCH 4/4] Raw Backend for UDST
2017-07-19 20:02 [Qemu-devel] Revised Unified Datagram Socket Transport patchset anton.ivanov
` (2 preceding siblings ...)
2017-07-19 20:02 ` [Qemu-devel] [PATCH 3/4] GRETAP Backend for UDST anton.ivanov
@ 2017-07-19 20:02 ` anton.ivanov
3 siblings, 0 replies; 5+ messages in thread
From: anton.ivanov @ 2017-07-19 20:02 UTC (permalink / raw)
To: qemu-devel; +Cc: jasowang, Anton Ivanov
From: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Raw Socket Backend for Universal Datagram Socket Transport
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
---
net/Makefile.objs | 2 +-
net/clients.h | 3 ++
net/net.c | 1 +
net/raw.c | 123 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
qapi-schema.json | 20 ++++++++-
qemu-options.hx | 32 ++++++++++++++
6 files changed, 178 insertions(+), 3 deletions(-)
create mode 100644 net/raw.c
diff --git a/net/Makefile.objs b/net/Makefile.objs
index 919bc3d78f..457297b5ed 100644
--- a/net/Makefile.objs
+++ b/net/Makefile.objs
@@ -2,7 +2,7 @@ common-obj-y = net.o queue.o checksum.o util.o hub.o
common-obj-y += socket.o
common-obj-y += dump.o
common-obj-y += eth.o
-common-obj-$(CONFIG_UDST) += udst.o l2tpv3.o gre.o
+common-obj-$(CONFIG_UDST) += udst.o l2tpv3.o gre.o raw.o
common-obj-$(CONFIG_POSIX) += vhost-user.o
common-obj-$(CONFIG_SLIRP) += slirp.o
common-obj-$(CONFIG_VDE) += vde.o
diff --git a/net/clients.h b/net/clients.h
index 8f8a59aee3..98d8ae59b7 100644
--- a/net/clients.h
+++ b/net/clients.h
@@ -53,6 +53,9 @@ int net_init_l2tpv3(const Netdev *netdev, const char *name,
int net_init_gre(const Netdev *netdev, const char *name,
NetClientState *peer, Error **errp);
+int net_init_raw(const Netdev *netdev, const char *name,
+ NetClientState *peer, Error **errp);
+
#ifdef CONFIG_VDE
int net_init_vde(const Netdev *netdev, const char *name,
NetClientState *peer, Error **errp);
diff --git a/net/net.c b/net/net.c
index 6163a8a3af..8eb0aa2bee 100644
--- a/net/net.c
+++ b/net/net.c
@@ -963,6 +963,7 @@ static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
#ifdef CONFIG_UDST
[NET_CLIENT_DRIVER_L2TPV3] = net_init_l2tpv3,
[NET_CLIENT_DRIVER_GRE] = net_init_gre,
+ [NET_CLIENT_DRIVER_RAW] = net_init_raw,
#endif
};
diff --git a/net/raw.c b/net/raw.c
new file mode 100644
index 0000000000..8f73248095
--- /dev/null
+++ b/net/raw.c
@@ -0,0 +1,123 @@
+/*
+ * QEMU System Emulator
+ *
+ * Copyright (c) 2015-2017 Cambridge Greys Limited
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ * Copyright (c) 2012-2014 Cisco Systems
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include <linux/ip.h>
+#include <netdb.h>
+#include <sys/ioctl.h>
+#include <net/if.h>
+#include "net/net.h"
+#include <sys/socket.h>
+#include <linux/if_packet.h>
+#include <net/ethernet.h>
+#include "clients.h"
+#include "qemu-common.h"
+#include "qemu/error-report.h"
+#include "qapi/error.h"
+#include "qemu/option.h"
+#include "qemu/sockets.h"
+#include "qemu/iov.h"
+#include "qemu/main-loop.h"
+#include "udst.h"
+
+static int noop(void *us, uint8_t *buf)
+{
+ return 0;
+}
+
+int net_init_raw(const Netdev *netdev,
+ const char *name,
+ NetClientState *peer, Error **errp)
+{
+
+ const NetdevRawOptions *raw;
+ NetUdstState *s;
+ NetClientState *nc;
+
+ int fd = -1;
+ int err;
+
+ struct ifreq ifr;
+ struct sockaddr_ll sock;
+
+
+ nc = qemu_new_udst_net_client(name, peer);
+
+ s = DO_UPCAST(NetUdstState, nc, nc);
+
+ fd = socket(AF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
+ if (fd == -1) {
+ err = -errno;
+ error_report("raw_open : raw socket creation failed, errno = %d", -err);
+ goto outerr;
+ }
+
+
+ s->dgram_dst = NULL;
+ s->dst_size = 0;
+
+ assert(netdev->type == NET_CLIENT_DRIVER_RAW);
+ raw = &netdev->u.raw;
+
+ memset(&ifr, 0, sizeof(struct ifreq));
+ strncpy((char *) &ifr.ifr_name, raw->ifname, sizeof(ifr.ifr_name) - 1);
+
+ if (ioctl(fd, SIOCGIFINDEX, (void *) &ifr) < 0) {
+ err = -errno;
+ error_report("SIOCGIFINDEX, failed to get raw interface index for %s",
+ raw->ifname);
+ goto outerr;
+ }
+
+ sock.sll_family = AF_PACKET;
+ sock.sll_protocol = htons(ETH_P_ALL);
+ sock.sll_ifindex = ifr.ifr_ifindex;
+
+ if (bind(fd, (struct sockaddr *) &sock, sizeof(struct sockaddr_ll)) < 0) {
+ error_report("raw: failed to bind raw socket");
+ err = -errno;
+ goto outerr;
+ }
+
+ s->offset = 0;
+
+ qemu_net_finalize_udst_init(s,
+ &noop,
+ NULL,
+ fd);
+
+ snprintf(s->nc.info_str, sizeof(s->nc.info_str),
+ "raw: connected");
+ return 0;
+outerr:
+ error_setg(errp, "Cannot initialize GRE transport");
+ qemu_del_net_client(nc);
+ if (fd >= 0) {
+ close(fd);
+ }
+ return -1;
+}
+
diff --git a/qapi-schema.json b/qapi-schema.json
index 082f56645a..8499c5403b 100644
--- a/qapi-schema.json
+++ b/qapi-schema.json
@@ -3900,6 +3900,21 @@
'data': { } }
##
+# @NetdevRawOptions:
+#
+# Connect the VLAN to an network interface using raw sockets
+#
+# @ifname: network interface name
+#
+
+# Since 2.9
+##
+{ 'struct': 'NetdevRawOptions',
+ 'data': {
+ 'ifname': 'str'
+} }
+
+##
# @NetdevVdeOptions:
#
# Connect the VLAN to a vde switch running on the host.
@@ -4017,7 +4032,7 @@
##
{ 'enum': 'NetClientDriver',
'data': [ 'none', 'nic', 'user', 'tap', 'l2tpv3', 'socket', 'vde', 'dump',
- 'bridge', 'hubport', 'netmap', 'vhost-user', 'udst', 'gre' ] }
+ 'bridge', 'hubport', 'netmap', 'vhost-user', 'udst', 'gre', 'raw' ] }
##
# @Netdev:
@@ -4051,7 +4066,8 @@
'netmap': 'NetdevNetmapOptions',
'vhost-user': 'NetdevVhostUserOptions',
'udst': 'NetdevUdstOptions',
- 'gre': 'NetdevGREOptions' } }
+ 'gre': 'NetdevGREOptions',
+ 'raw': 'NetdevRawOptions' } }
##
# @NetLegacy:
diff --git a/qemu-options.hx b/qemu-options.hx
index e35ef032cf..12d2538db8 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -1988,6 +1988,13 @@ DEF("netdev", HAS_ARG, QEMU_OPTION_netdev,
" use 'txkey=0x01234' to specify a txkey\n"
" use 'sequence=on' to add frame sequence to each packet\n"
" use 'pinsequence=on' to work around broken sequence handling in peer\n"
+ "-netdev raw,id=str,ifname=ifname\n"
+ " configure a network backend with ID 'str' connected to\n"
+ " an Ethernet interface named ifname via raw socket.\n"
+ " This backend does not change the interface settings.\n"
+ " Most interfaces will require being set into promisc mode,\n"
+ " as well having most offloads (TSO, etc) turned off.\n"
+ " Some virtual interfaces like tap support only RX.\n"
#endif
"-netdev socket,id=str[,fd=h][,listen=[host]:port][,connect=host:port]\n"
" configure a network backend to connect to another network\n"
@@ -2462,6 +2469,31 @@ qemu-system-i386 linux.img -device virtio-net-pci,netdev=gre0 -netdev gre,id=gre
@end example
+@item -netdev raw,id=@var{id},ifname=@var{ifname}
+Connect VLAN @var{n} directly to an Ethernet interface using raw socket.
+
+This transport allows a VM to bypass most of the network stack which is
+extremely useful for tapping.
+
+@item ifname=@var{ifname}
+ interface name (mandatory)
+
+@example
+# set up the interface - put it in promiscuous mode and turn off offloads
+ifconfig eth0 up
+ifconfig eth0 promisc
+
+/sbin/ethtool -K eth0 gro off
+/sbin/ethtool -K eth0 tso off
+/sbin/ethtool -K eth0 gso off
+/sbin/ethtool -K eth0 tx off
+
+# launch QEMU instance - if your network has reorder or is very lossy add ,pincounter
+
+qemu-system-i386 linux.img -device virtio-net-pci,netdev=raw0 -netdev raw,id=raw0,ifname=eth0
+
+@end example
+
@item -netdev vde,id=@var{id}[,sock=@var{socketpath}][,port=@var{n}][,group=@var{groupname}][,mode=@var{octalmode}]
@itemx -net vde[,vlan=@var{n}][,name=@var{name}][,sock=@var{socketpath}] [,port=@var{n}][,group=@var{groupname}][,mode=@var{octalmode}]
Connect VLAN @var{n} to PORT @var{n} of a vde switch running on host and
--
2.11.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2017-07-19 20:03 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-07-19 20:02 [Qemu-devel] Revised Unified Datagram Socket Transport patchset anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 1/4] Unified Datagram Socket Transports anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 2/4] Migrate l2tpv3 to UDST Backend anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 3/4] GRETAP Backend for UDST anton.ivanov
2017-07-19 20:02 ` [Qemu-devel] [PATCH 4/4] Raw " anton.ivanov
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.