* [userspace meter v3 0/5] Userspace meter
@ 2017-01-31 6:39 Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 1/5] netdev-dummy: Add --len option for netdev-dummy/receive command Andy Zhou
` (5 more replies)
0 siblings, 6 replies; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 6:39 UTC (permalink / raw)
To: davem; +Cc: netdev, Andy Zhou
Repost user space meter support. This is based Jarno's original work
at: https://mail.openvswitch.org/pipermail/ovs-dev/2015-November/306304.html.
With some enhancements, and rebased to current master.
---
v1-v2: rebase and repost.
v2-v3: simplify patch 2/5.
Andy Zhou (2):
netdev-dummy: Add --len option for netdev-dummy/receive command
dp-packet: Enhance packet batch APIs.
Jarno Rajahalme (3):
dpif: Meter framework.
ofproto: Meter translation.
dpif-netdev: Simple DROP meter implementation.
datapath/linux/compat/include/linux/openvswitch.h | 4 +-
include/openvswitch/ofp-actions.h | 1 +
lib/dp-packet.h | 140 +++++--
lib/dpif-netdev.c | 433 ++++++++++++++++++++--
lib/dpif-netlink.c | 46 ++-
lib/dpif-provider.h | 29 ++
lib/dpif.c | 128 ++++++-
lib/dpif.h | 13 +-
lib/netdev-dummy.c | 48 ++-
lib/netdev-linux.c | 3 +-
lib/netdev.c | 24 +-
lib/odp-execute.c | 58 +--
lib/odp-util.c | 14 +
lib/ofp-actions.c | 1 +
ofproto/ofproto-dpif-sflow.c | 1 +
ofproto/ofproto-dpif-xlate.c | 13 +-
ofproto/ofproto-dpif.c | 60 ++-
ofproto/ofproto-provider.h | 13 +-
ofproto/ofproto.c | 51 +--
tests/dpif-netdev.at | 106 ++++++
tests/test-conntrack.c | 56 ++-
21 files changed, 1036 insertions(+), 206 deletions(-)
--
1.9.1
^ permalink raw reply [flat|nested] 8+ messages in thread
* [userspace meter v3 1/5] netdev-dummy: Add --len option for netdev-dummy/receive command
2017-01-31 6:39 [userspace meter v3 0/5] Userspace meter Andy Zhou
@ 2017-01-31 6:39 ` Andy Zhou
2017-01-31 7:30 ` Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 2/5] dp-packet: Enhance packet batch APIs Andy Zhou
` (4 subsequent siblings)
5 siblings, 1 reply; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 6:39 UTC (permalink / raw)
To: davem; +Cc: netdev, Andy Zhou
Currently, there is no way to specify the packet size when injecting
a packet via "netdev-dummy/receive" with a flow specification. Thus
far, packet size is not important for testing OVS features, but it
becomes useful in writing unit tests for the meter implementation
in a later patch.
Signed-off-by: Andy Zhou <azhou@ovn.org>
---
lib/netdev-dummy.c | 38 ++++++++++++++++++++++++++++----------
1 file changed, 28 insertions(+), 10 deletions(-)
diff --git a/lib/netdev-dummy.c b/lib/netdev-dummy.c
index e6e36cd..10df0a7 100644
--- a/lib/netdev-dummy.c
+++ b/lib/netdev-dummy.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2010, 2011, 2012, 2013, 2015, 2016 Nicira, Inc.
+ * Copyright (c) 2010, 2011, 2012, 2013, 2015, 2016, 2017 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@@ -1433,7 +1433,15 @@ pkt_list_delete(struct ovs_list *l)
}
static struct dp_packet *
-eth_from_packet_or_flow(const char *s)
+eth_from_packet(const char *s)
+{
+ struct dp_packet *packet;
+ eth_from_hex(s, &packet);
+ return packet;
+}
+
+static struct dp_packet *
+eth_from_flow(const char *s)
{
enum odp_key_fitness fitness;
struct dp_packet *packet;
@@ -1441,10 +1449,6 @@ eth_from_packet_or_flow(const char *s)
struct flow flow;
int error;
- if (!eth_from_hex(s, &packet)) {
- return packet;
- }
-
/* Convert string to datapath key.
*
* It would actually be nicer to parse an OpenFlow-like flow key here, but
@@ -1540,10 +1544,24 @@ netdev_dummy_receive(struct unixctl_conn *conn,
for (i = k; i < argc; i++) {
struct dp_packet *packet;
- packet = eth_from_packet_or_flow(argv[i]);
+ /* Try to parse 'argv[i]' as packet in hex. */
+ packet = eth_from_packet(argv[i]);
+
if (!packet) {
- unixctl_command_reply_error(conn, "bad packet syntax");
- goto exit;
+ /* Try parse 'argv[i]' as odp flow. */
+ packet = eth_from_flow(argv[i]);
+
+ if (!packet) {
+ unixctl_command_reply_error(conn, "bad packet or flow syntax");
+ goto exit;
+ }
+
+ /* Parse optional --len argument immediately follow a 'flow'. */
+ if (argc >= i + 2 && !strcmp(argv[i + 1], "--len")) {
+ int packet_size = strtol(argv[i + 2], NULL, 10);
+ dp_packet_set_size(packet, packet_size);
+ i+=2;
+ }
}
netdev_dummy_queue_packet(dummy_dev, packet, rx_qid);
@@ -1757,7 +1775,7 @@ void
netdev_dummy_register(enum dummy_level level)
{
unixctl_command_register("netdev-dummy/receive",
- "name [--qid queue_id] packet|flow...",
+ "name [--qid queue_id] packet|flow [--len packet len]..",
2, INT_MAX, netdev_dummy_receive, NULL);
unixctl_command_register("netdev-dummy/set-admin-state",
"[netdev] up|down", 1, 2,
--
1.9.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [userspace meter v3 2/5] dp-packet: Enhance packet batch APIs.
2017-01-31 6:39 [userspace meter v3 0/5] Userspace meter Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 1/5] netdev-dummy: Add --len option for netdev-dummy/receive command Andy Zhou
@ 2017-01-31 6:39 ` Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 3/5] dpif: Meter framework Andy Zhou
` (3 subsequent siblings)
5 siblings, 0 replies; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 6:39 UTC (permalink / raw)
To: davem; +Cc: netdev, Andy Zhou
One common use case of 'struct dp_packet_batch' is to process all
packets in the batch in order. Add an iterator for this use case
to simplify the logic of calling sites,
Another common use case is to drop packets in the batch, by reading
all packets, but writing back pointers of fewer packets. Add macros
to support this use case.
Signed-off-by: Andy Zhou <azhou@ovn.org>
---
lib/dp-packet.h | 140 +++++++++++++++++++++++++++++++------------
lib/dpif-netdev.c | 62 +++++++++----------
lib/dpif.c | 2 +-
lib/netdev-dummy.c | 10 ++--
lib/netdev-linux.c | 3 +-
lib/netdev.c | 24 ++++----
lib/odp-execute.c | 55 ++++++++---------
ofproto/ofproto-dpif-xlate.c | 2 +-
tests/test-conntrack.c | 56 +++++++----------
9 files changed, 201 insertions(+), 153 deletions(-)
diff --git a/lib/dp-packet.h b/lib/dp-packet.h
index cf7d247..d0c14a5 100644
--- a/lib/dp-packet.h
+++ b/lib/dp-packet.h
@@ -632,79 +632,143 @@ reset_dp_packet_checksum_ol_flags(struct dp_packet *p)
enum { NETDEV_MAX_BURST = 32 }; /* Maximum number packets in a batch. */
struct dp_packet_batch {
- int count;
+ size_t count;
bool trunc; /* true if the batch needs truncate. */
struct dp_packet *packets[NETDEV_MAX_BURST];
};
-static inline void dp_packet_batch_init(struct dp_packet_batch *b)
+static inline void
+dp_packet_batch_init(struct dp_packet_batch *batch)
{
- b->count = 0;
- b->trunc = false;
+ batch->count = 0;
+ batch->trunc = false;
}
static inline void
-dp_packet_batch_clone(struct dp_packet_batch *dst,
- struct dp_packet_batch *src)
+dp_packet_batch_add__(struct dp_packet_batch *batch,
+ struct dp_packet *packet, size_t limit)
{
- int i;
-
- for (i = 0; i < src->count; i++) {
- dst->packets[i] = dp_packet_clone(src->packets[i]);
+ if (batch->count < limit) {
+ batch->packets[batch->count++] = packet;
+ } else {
+ dp_packet_delete(packet);
}
- dst->count = src->count;
- dst->trunc = src->trunc;
}
static inline void
-packet_batch_init_packet(struct dp_packet_batch *b, struct dp_packet *p)
+dp_packet_batch_add(struct dp_packet_batch *batch, struct dp_packet *packet)
+{
+ dp_packet_batch_add__(batch, packet, NETDEV_MAX_BURST);
+}
+
+static inline size_t
+dp_packet_batch_size(const struct dp_packet_batch *batch)
+{
+ return batch->count;
+}
+
+/*
+ * Clear 'batch' for refill. Use dp_packet_batch_refill() to add
+ * packets back into the 'batch'.
+ *
+ * Return the original size of the 'batch'. */
+static inline void
+dp_packet_batch_refill_init(struct dp_packet_batch *batch)
{
- b->count = 1;
- b->trunc = false;
- b->packets[0] = p;
+ batch->count = 0;
+};
+
+static inline void
+dp_packet_batch_refill(struct dp_packet_batch *batch,
+ struct dp_packet *packet, size_t idx)
+{
+ dp_packet_batch_add__(batch, packet, MIN(NETDEV_MAX_BURST, idx + 1));
+}
+
+static inline void
+dp_packet_batch_init_packet(struct dp_packet_batch *batch, struct dp_packet *p)
+{
+ dp_packet_batch_init(batch);
+ batch->count = 1;
+ batch->packets[0] = p;
+}
+
+static inline bool
+dp_packet_batch_is_empty(const struct dp_packet_batch *batch)
+{
+ return !dp_packet_batch_size(batch);
+}
+
+#define DP_PACKET_BATCH_FOR_EACH(PACKET, BATCH) \
+ for (size_t i = 0; i < dp_packet_batch_size(BATCH); i++) \
+ if ((PACKET = BATCH->packets[i]) != NULL)
+
+/* Use this macro for cases where some packets in the 'BATCH' may be
+ * dropped after going through each packet in the 'BATCH'.
+ *
+ * For packets to stay in the 'BATCH', they need to be refilled back
+ * into the 'BATCH' by calling dp_packet_batch_refill(). Caller owns
+ * the packets that are not refilled.
+ *
+ * Caller needs to supply 'SIZE', that stores the current number of
+ * packets in 'BATCH'. It is best to declare this variable with
+ * the 'const' modifier since it should not be modified by
+ * the iterator. */
+#define DP_PACKET_BATCH_REFILL_FOR_EACH(IDX, SIZE, PACKET, BATCH) \
+ for (dp_packet_batch_refill_init(BATCH), IDX=0; IDX < SIZE; IDX++) \
+ if ((PACKET = BATCH->packets[IDX]) != NULL)
+
+static inline void
+dp_packet_batch_clone(struct dp_packet_batch *dst,
+ struct dp_packet_batch *src)
+{
+ struct dp_packet *packet;
+
+ dp_packet_batch_init(dst);
+ DP_PACKET_BATCH_FOR_EACH (packet, src) {
+ dp_packet_batch_add(dst, dp_packet_clone(packet));
+ }
}
static inline void
dp_packet_delete_batch(struct dp_packet_batch *batch, bool may_steal)
{
if (may_steal) {
- int i;
+ struct dp_packet *packet;
- for (i = 0; i < batch->count; i++) {
- dp_packet_delete(batch->packets[i]);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ dp_packet_delete(packet);
}
+ dp_packet_batch_init(batch);
}
}
static inline void
-dp_packet_batch_apply_cutlen(struct dp_packet_batch *pktb)
+dp_packet_batch_apply_cutlen(struct dp_packet_batch *batch)
{
- int i;
-
- if (!pktb->trunc)
- return;
+ if (batch->trunc) {
+ struct dp_packet *packet;
- for (i = 0; i < pktb->count; i++) {
- uint32_t cutlen = dp_packet_get_cutlen(pktb->packets[i]);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ uint32_t cutlen = dp_packet_get_cutlen(packet);
- dp_packet_set_size(pktb->packets[i],
- dp_packet_size(pktb->packets[i]) - cutlen);
- dp_packet_reset_cutlen(pktb->packets[i]);
+ dp_packet_set_size(packet, dp_packet_size(packet) - cutlen);
+ dp_packet_reset_cutlen(packet);
+ }
+ batch->trunc = false;
}
- pktb->trunc = false;
}
static inline void
-dp_packet_batch_reset_cutlen(struct dp_packet_batch *pktb)
+dp_packet_batch_reset_cutlen(struct dp_packet_batch *batch)
{
- int i;
+ if (batch->trunc) {
+ struct dp_packet *packet;
- if (!pktb->trunc)
- return;
-
- pktb->trunc = false;
- for (i = 0; i < pktb->count; i++) {
- dp_packet_reset_cutlen(pktb->packets[i]);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ dp_packet_reset_cutlen(packet);
+ }
+ batch->trunc = false;
}
}
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index 42631bc..719a518 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -2683,7 +2683,7 @@ dpif_netdev_execute(struct dpif *dpif, struct dpif_execute *execute)
flow_hash_5tuple(execute->flow, 0));
}
- packet_batch_init_packet(&pp, execute->packet);
+ dp_packet_batch_init_packet(&pp, execute->packet);
dp_netdev_execute_actions(pmd, &pp, false, execute->flow,
execute->actions, execute->actions_len,
time_msec());
@@ -4073,20 +4073,21 @@ dp_netdev_queue_batches(struct dp_packet *pkt,
* initialized by this function using 'port_no'.
*/
static inline size_t
-emc_processing(struct dp_netdev_pmd_thread *pmd, struct dp_packet_batch *packets_,
+emc_processing(struct dp_netdev_pmd_thread *pmd,
+ struct dp_packet_batch *packets_,
struct netdev_flow_key *keys,
struct packet_batch_per_flow batches[], size_t *n_batches,
bool md_is_valid, odp_port_t port_no)
{
struct emc_cache *flow_cache = &pmd->flow_cache;
struct netdev_flow_key *key = &keys[0];
- size_t i, n_missed = 0, n_dropped = 0;
- struct dp_packet **packets = packets_->packets;
- int cnt = packets_->count;
+ size_t n_missed = 0, n_dropped = 0;
+ struct dp_packet *packet;
+ const size_t size = dp_packet_batch_size(packets_);
+ int i;
- for (i = 0; i < cnt; i++) {
+ DP_PACKET_BATCH_REFILL_FOR_EACH (i, size, packet, packets_) {
struct dp_netdev_flow *flow;
- struct dp_packet *packet = packets[i];
if (OVS_UNLIKELY(dp_packet_size(packet) < ETH_HEADER_LEN)) {
dp_packet_delete(packet);
@@ -4094,7 +4095,8 @@ emc_processing(struct dp_netdev_pmd_thread *pmd, struct dp_packet_batch *packets
continue;
}
- if (i != cnt - 1) {
+ if (i != size - 1) {
+ struct dp_packet **packets = packets_->packets;
/* Prefetch next packet data and metadata. */
OVS_PREFETCH(dp_packet_data(packets[i+1]));
pkt_metadata_prefetch_init(&packets[i+1]->md);
@@ -4113,8 +4115,8 @@ emc_processing(struct dp_netdev_pmd_thread *pmd, struct dp_packet_batch *packets
n_batches);
} else {
/* Exact match cache missed. Group missed packets together at
- * the beginning of the 'packets' array. */
- packets[n_missed] = packet;
+ * the beginning of the 'packets' array. */
+ dp_packet_batch_refill(packets_, packet, i);
/* 'key[n_missed]' contains the key of the current packet and it
* must be returned to the caller. The next key should be extracted
* to 'keys[n_missed + 1]'. */
@@ -4122,9 +4124,9 @@ emc_processing(struct dp_netdev_pmd_thread *pmd, struct dp_packet_batch *packets
}
}
- dp_netdev_count_packet(pmd, DP_STAT_EXACT_HIT, cnt - n_dropped - n_missed);
+ dp_netdev_count_packet(pmd, DP_STAT_EXACT_HIT, size - n_dropped - n_missed);
- return n_missed;
+ return dp_packet_batch_size(packets_);
}
static inline void
@@ -4168,7 +4170,7 @@ handle_packet_upcall(struct dp_netdev_pmd_thread *pmd, struct dp_packet *packet,
/* We can't allow the packet batching in the next loop to execute
* the actions. Otherwise, if there are any slow path actions,
* we'll send the packet up twice. */
- packet_batch_init_packet(&b, packet);
+ dp_packet_batch_init_packet(&b, packet);
dp_netdev_execute_actions(pmd, &b, true, &match.flow,
actions->data, actions->size, now);
@@ -4316,14 +4318,13 @@ dp_netdev_input__(struct dp_netdev_pmd_thread *pmd,
OVS_ALIGNED_VAR(CACHE_LINE_SIZE) struct netdev_flow_key keys[PKT_ARRAY_SIZE];
struct packet_batch_per_flow batches[PKT_ARRAY_SIZE];
long long now = time_msec();
- size_t newcnt, n_batches, i;
+ size_t n_batches;
odp_port_t in_port;
n_batches = 0;
- newcnt = emc_processing(pmd, packets, keys, batches, &n_batches,
+ emc_processing(pmd, packets, keys, batches, &n_batches,
md_is_valid, port_no);
- if (OVS_UNLIKELY(newcnt)) {
- packets->count = newcnt;
+ if (!dp_packet_batch_is_empty(packets)) {
/* Get ingress port from first packet's metadata. */
in_port = packets->packets[0]->md.in_port.odp_port;
fast_path_processing(pmd, packets, keys, batches, &n_batches, in_port, now);
@@ -4338,6 +4339,7 @@ dp_netdev_input__(struct dp_netdev_pmd_thread *pmd,
* already its own batches[k] still waiting to be served. So if its
* ‘batch’ member is not reset, the recirculated packet would be wrongly
* appended to batches[k] of the 1st call to dp_netdev_input__(). */
+ size_t i;
for (i = 0; i < n_batches; i++) {
batches[i].flow->batch = NULL;
}
@@ -4512,7 +4514,7 @@ dp_execute_userspace_action(struct dp_netdev_pmd_thread *pmd,
DPIF_UC_ACTION, userdata, actions,
NULL);
if (!error || error == ENOSPC) {
- packet_batch_init_packet(&b, packet);
+ dp_packet_batch_init_packet(&b, packet);
dp_netdev_execute_actions(pmd, &b, may_steal, flow,
actions->data, actions->size, now);
} else if (may_steal) {
@@ -4584,7 +4586,6 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
p = pmd_tnl_port_cache_lookup(pmd, portno);
if (p) {
struct dp_packet_batch tnl_pkt;
- int i;
if (!may_steal) {
dp_packet_batch_clone(&tnl_pkt, packets_);
@@ -4595,12 +4596,13 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
dp_packet_batch_apply_cutlen(packets_);
netdev_pop_header(p->port->netdev, packets_);
- if (!packets_->count) {
+ if (dp_packet_batch_is_empty(packets_)) {
return;
}
- for (i = 0; i < packets_->count; i++) {
- packets_->packets[i]->md.in_port.odp_port = portno;
+ struct dp_packet *packet;
+ DP_PACKET_BATCH_FOR_EACH (packet, packets_) {
+ packet->md.in_port.odp_port = portno;
}
(*depth)++;
@@ -4614,14 +4616,12 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
case OVS_ACTION_ATTR_USERSPACE:
if (!fat_rwlock_tryrdlock(&dp->upcall_rwlock)) {
struct dp_packet_batch *orig_packets_ = packets_;
- struct dp_packet **packets = packets_->packets;
const struct nlattr *userdata;
struct dp_packet_batch usr_pkt;
struct ofpbuf actions;
struct flow flow;
ovs_u128 ufid;
bool clone = false;
- int i;
userdata = nl_attr_find_nested(a, OVS_USERSPACE_ATTR_USERDATA);
ofpbuf_init(&actions, 0);
@@ -4630,7 +4630,6 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
if (!may_steal) {
dp_packet_batch_clone(&usr_pkt, packets_);
packets_ = &usr_pkt;
- packets = packets_->packets;
clone = true;
dp_packet_batch_reset_cutlen(orig_packets_);
}
@@ -4638,10 +4637,11 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
dp_packet_batch_apply_cutlen(packets_);
}
- for (i = 0; i < packets_->count; i++) {
- flow_extract(packets[i], &flow);
+ struct dp_packet *packet;
+ DP_PACKET_BATCH_FOR_EACH (packet, packets_) {
+ flow_extract(packet, &flow);
dpif_flow_hash(dp->dpif, &flow, sizeof flow, &ufid);
- dp_execute_userspace_action(pmd, packets[i], may_steal, &flow,
+ dp_execute_userspace_action(pmd, packet, may_steal, &flow,
&ufid, &actions, userdata, now);
}
@@ -4659,15 +4659,15 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
case OVS_ACTION_ATTR_RECIRC:
if (*depth < MAX_RECIRC_DEPTH) {
struct dp_packet_batch recirc_pkts;
- int i;
if (!may_steal) {
dp_packet_batch_clone(&recirc_pkts, packets_);
packets_ = &recirc_pkts;
}
- for (i = 0; i < packets_->count; i++) {
- packets_->packets[i]->md.recirc_id = nl_attr_get_u32(a);
+ struct dp_packet *packet;
+ DP_PACKET_BATCH_FOR_EACH (packet, packets_) {
+ packet->md.recirc_id = nl_attr_get_u32(a);
}
(*depth)++;
diff --git a/lib/dpif.c b/lib/dpif.c
index 7c953b5..374f013 100644
--- a/lib/dpif.c
+++ b/lib/dpif.c
@@ -1202,7 +1202,7 @@ dpif_execute_with_help(struct dpif *dpif, struct dpif_execute *execute)
COVERAGE_INC(dpif_execute_with_help);
- packet_batch_init_packet(&pb, execute->packet);
+ dp_packet_batch_init_packet(&pb, execute->packet);
odp_execute_actions(&aux, &pb, false, execute->actions,
execute->actions_len, dpif_execute_helper_cb);
return aux.error;
diff --git a/lib/netdev-dummy.c b/lib/netdev-dummy.c
index 10df0a7..e02beae 100644
--- a/lib/netdev-dummy.c
+++ b/lib/netdev-dummy.c
@@ -1061,13 +1061,13 @@ netdev_dummy_send(struct netdev *netdev, int qid OVS_UNUSED,
{
struct netdev_dummy *dev = netdev_dummy_cast(netdev);
int error = 0;
- int i;
- for (i = 0; i < batch->count; i++) {
- const void *buffer = dp_packet_data(batch->packets[i]);
- size_t size = dp_packet_size(batch->packets[i]);
+ struct dp_packet *packet;
+ DP_PACKET_BATCH_FOR_EACH(packet, batch) {
+ const void *buffer = dp_packet_data(packet);
+ size_t size = dp_packet_size(packet);
- size -= dp_packet_get_cutlen(batch->packets[i]);
+ size -= dp_packet_get_cutlen(packet);
if (size < ETH_HEADER_LEN) {
error = EMSGSIZE;
diff --git a/lib/netdev-linux.c b/lib/netdev-linux.c
index a5a9ec1..9ff1333 100644
--- a/lib/netdev-linux.c
+++ b/lib/netdev-linux.c
@@ -1125,8 +1125,7 @@ netdev_linux_rxq_recv(struct netdev_rxq *rxq_, struct dp_packet_batch *batch)
}
dp_packet_delete(buffer);
} else {
- batch->packets[0] = buffer;
- batch->count = 1;
+ dp_packet_batch_init_packet(batch, buffer);
}
return retval;
diff --git a/lib/netdev.c b/lib/netdev.c
index 839b1f6..1e6bb2b 100644
--- a/lib/netdev.c
+++ b/lib/netdev.c
@@ -749,20 +749,19 @@ netdev_send(struct netdev *netdev, int qid, struct dp_packet_batch *batch,
void
netdev_pop_header(struct netdev *netdev, struct dp_packet_batch *batch)
{
- int i, n_cnt = 0;
- struct dp_packet **buffers = batch->packets;
+ struct dp_packet *packet;
+ size_t i, size = dp_packet_batch_size(batch);
- for (i = 0; i < batch->count; i++) {
- buffers[i] = netdev->netdev_class->pop_header(buffers[i]);
- if (buffers[i]) {
+ DP_PACKET_BATCH_REFILL_FOR_EACH (i, size, packet, batch) {
+ packet = netdev->netdev_class->pop_header(packet);
+ if (packet) {
/* Reset the checksum offload flags if present, to avoid wrong
* interpretation in the further packet processing when
* recirculated.*/
- reset_dp_packet_checksum_ol_flags(buffers[i]);
- buffers[n_cnt++] = buffers[i];
+ reset_dp_packet_checksum_ol_flags(packet);
+ dp_packet_batch_refill(batch, packet, i);
}
}
- batch->count = n_cnt;
}
void
@@ -799,11 +798,10 @@ netdev_push_header(const struct netdev *netdev,
struct dp_packet_batch *batch,
const struct ovs_action_push_tnl *data)
{
- int i;
-
- for (i = 0; i < batch->count; i++) {
- netdev->netdev_class->push_header(batch->packets[i], data);
- pkt_metadata_init(&batch->packets[i]->md, u32_to_odp(data->out_port));
+ struct dp_packet *packet;
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ netdev->netdev_class->push_header(packet, data);
+ pkt_metadata_init(&packet->md, u32_to_odp(data->out_port));
}
return 0;
diff --git a/lib/odp-execute.c b/lib/odp-execute.c
index c4ae5ce..465280b 100644
--- a/lib/odp-execute.c
+++ b/lib/odp-execute.c
@@ -523,7 +523,7 @@ odp_execute_sample(void *dp, struct dp_packet *packet, bool steal,
}
}
- packet_batch_init_packet(&pb, packet);
+ dp_packet_batch_init_packet(&pb, packet);
odp_execute_actions(dp, &pb, steal, nl_attr_get(subactions),
nl_attr_get_size(subactions), dp_execute_action);
}
@@ -543,7 +543,7 @@ odp_execute_clone(void *dp, struct dp_packet *packet, bool steal,
* will free the clone. */
packet = dp_packet_clone(packet);
}
- packet_batch_init_packet(&pb, packet);
+ dp_packet_batch_init_packet(&pb, packet);
odp_execute_actions(dp, &pb, true, nl_attr_get(actions),
nl_attr_get_size(actions), dp_execute_action);
}
@@ -588,11 +588,9 @@ odp_execute_actions(void *dp, struct dp_packet_batch *batch, bool steal,
const struct nlattr *actions, size_t actions_len,
odp_execute_cb dp_execute_action)
{
- struct dp_packet **packets = batch->packets;
- int cnt = batch->count;
+ struct dp_packet *packet;
const struct nlattr *a;
unsigned int left;
- int i;
NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
int type = nl_attr_type(a);
@@ -627,11 +625,10 @@ odp_execute_actions(void *dp, struct dp_packet_batch *batch, bool steal,
struct flow flow;
uint32_t hash;
- for (i = 0; i < cnt; i++) {
- flow_extract(packets[i], &flow);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ flow_extract(packet, &flow);
hash = flow_hash_5tuple(&flow, hash_act->hash_basis);
-
- packets[i]->md.dp_hash = hash;
+ packet->md.dp_hash = hash;
}
} else {
/* Assert on unknown hash algorithm. */
@@ -643,48 +640,48 @@ odp_execute_actions(void *dp, struct dp_packet_batch *batch, bool steal,
case OVS_ACTION_ATTR_PUSH_VLAN: {
const struct ovs_action_push_vlan *vlan = nl_attr_get(a);
- for (i = 0; i < cnt; i++) {
- eth_push_vlan(packets[i], vlan->vlan_tpid, vlan->vlan_tci);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ eth_push_vlan(packet, vlan->vlan_tpid, vlan->vlan_tci);
}
break;
}
case OVS_ACTION_ATTR_POP_VLAN:
- for (i = 0; i < cnt; i++) {
- eth_pop_vlan(packets[i]);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ eth_pop_vlan(packet);
}
break;
case OVS_ACTION_ATTR_PUSH_MPLS: {
const struct ovs_action_push_mpls *mpls = nl_attr_get(a);
- for (i = 0; i < cnt; i++) {
- push_mpls(packets[i], mpls->mpls_ethertype, mpls->mpls_lse);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ push_mpls(packet, mpls->mpls_ethertype, mpls->mpls_lse);
}
break;
}
case OVS_ACTION_ATTR_POP_MPLS:
- for (i = 0; i < cnt; i++) {
- pop_mpls(packets[i], nl_attr_get_be16(a));
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ pop_mpls(packet, nl_attr_get_be16(a));
}
break;
case OVS_ACTION_ATTR_SET:
- for (i = 0; i < cnt; i++) {
- odp_execute_set_action(packets[i], nl_attr_get(a));
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ odp_execute_set_action(packet, nl_attr_get(a));
}
break;
case OVS_ACTION_ATTR_SET_MASKED:
- for (i = 0; i < cnt; i++) {
- odp_execute_masked_set_action(packets[i], nl_attr_get(a));
+ DP_PACKET_BATCH_FOR_EACH(packet, batch) {
+ odp_execute_masked_set_action(packet, nl_attr_get(a));
}
break;
case OVS_ACTION_ATTR_SAMPLE:
- for (i = 0; i < cnt; i++) {
- odp_execute_sample(dp, packets[i], steal && last_action, a,
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ odp_execute_sample(dp, packet, steal && last_action, a,
dp_execute_action);
}
@@ -700,15 +697,15 @@ odp_execute_actions(void *dp, struct dp_packet_batch *batch, bool steal,
nl_attr_get_unspec(a, sizeof *trunc);
batch->trunc = true;
- for (i = 0; i < cnt; i++) {
- dp_packet_set_cutlen(packets[i], trunc->max_len);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ dp_packet_set_cutlen(packet, trunc->max_len);
}
break;
}
case OVS_ACTION_ATTR_CLONE:
- for (i = 0; i < cnt; i++) {
- odp_execute_clone(dp, packets[i], steal && last_action, a,
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ odp_execute_clone(dp, packet, steal && last_action, a,
dp_execute_action);
}
@@ -732,8 +729,8 @@ odp_execute_actions(void *dp, struct dp_packet_batch *batch, bool steal,
}
if (steal) {
- for (i = 0; i < cnt; i++) {
- dp_packet_delete(packets[i]);
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
+ dp_packet_delete(packet);
}
}
}
diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c
index 0513394..48b27a6 100644
--- a/ofproto/ofproto-dpif-xlate.c
+++ b/ofproto/ofproto-dpif-xlate.c
@@ -3825,7 +3825,7 @@ execute_controller_action(struct xlate_ctx *ctx, int len,
}
packet = dp_packet_clone(ctx->xin->packet);
- packet_batch_init_packet(&batch, packet);
+ dp_packet_batch_init_packet(&batch, packet);
odp_execute_actions(NULL, &batch, false,
ctx->odp_actions->data, ctx->odp_actions->size, NULL);
diff --git a/tests/test-conntrack.c b/tests/test-conntrack.c
index 803e2b9..e362f8a 100644
--- a/tests/test-conntrack.c
+++ b/tests/test-conntrack.c
@@ -39,8 +39,6 @@ prepare_packets(size_t n, bool change, unsigned tid, ovs_be16 *dl_type)
ovs_assert(n <= ARRAY_SIZE(pkt_batch->packets));
dp_packet_batch_init(pkt_batch);
- pkt_batch->count = n;
-
for (i = 0; i < n; i++) {
struct udp_header *udp;
struct dp_packet *pkt = dp_packet_new(sizeof payload/2);
@@ -55,11 +53,10 @@ prepare_packets(size_t n, bool change, unsigned tid, ovs_be16 *dl_type)
udp->udp_dst = htons(ntohs(udp->udp_dst) + i);
}
- pkt_batch->packets[i] = pkt;
+ dp_packet_batch_add(pkt_batch, pkt);
*dl_type = flow.dl_type;
}
-
return pkt_batch;
}
@@ -154,7 +151,6 @@ static void
pcap_batch_execute_conntrack(struct conntrack *ct,
struct dp_packet_batch *pkt_batch)
{
- size_t i;
struct dp_packet_batch new_batch;
ovs_be16 dl_type = htons(0);
@@ -162,15 +158,14 @@ pcap_batch_execute_conntrack(struct conntrack *ct,
/* pkt_batch contains packets with different 'dl_type'. We have to
* call conntrack_execute() on packets with the same 'dl_type'. */
-
- for (i = 0; i < pkt_batch->count; i++) {
- struct dp_packet *pkt = pkt_batch->packets[i];
+ struct dp_packet *packet;
+ DP_PACKET_BATCH_FOR_EACH (packet, pkt_batch) {
struct flow flow;
/* This also initializes the l3 and l4 pointers. */
- flow_extract(pkt, &flow);
+ flow_extract(packet, &flow);
- if (!new_batch.count) {
+ if (dp_packet_batch_is_empty(&new_batch)) {
dl_type = flow.dl_type;
}
@@ -179,10 +174,10 @@ pcap_batch_execute_conntrack(struct conntrack *ct,
NULL);
dp_packet_batch_init(&new_batch);
}
- new_batch.packets[new_batch.count++] = pkt;
+ new_batch.packets[new_batch.count++] = packet;;
}
- if (new_batch.count) {
+ if (!dp_packet_batch_is_empty(&new_batch)) {
conntrack_execute(ct, &new_batch, dl_type, true, 0, NULL, NULL, NULL);
}
@@ -191,9 +186,9 @@ pcap_batch_execute_conntrack(struct conntrack *ct,
static void
test_pcap(struct ovs_cmdl_context *ctx)
{
- size_t total_count, i, batch_size;
+ size_t total_count, batch_size;
FILE *pcap;
- int err;
+ int err = 0;
pcap = ovs_pcap_open(ctx->argv[1], "rb");
if (!pcap) {
@@ -214,41 +209,36 @@ test_pcap(struct ovs_cmdl_context *ctx)
conntrack_init(&ct);
total_count = 0;
- for (;;) {
- struct dp_packet_batch pkt_batch;
-
- dp_packet_batch_init(&pkt_batch);
+ while (!err) {
+ struct dp_packet *packet;
+ struct dp_packet_batch pkt_batch_;
+ struct dp_packet_batch *batch = &pkt_batch_;
- for (i = 0; i < batch_size; i++) {
- err = ovs_pcap_read(pcap, &pkt_batch.packets[i], NULL);
- if (err) {
- break;
- }
+ dp_packet_batch_init(batch);
+ err = ovs_pcap_read(pcap, &packet, NULL);
+ if (err) {
+ break;
}
+ dp_packet_batch_add(batch, packet);
- pkt_batch.count = i;
- if (pkt_batch.count == 0) {
+ if (dp_packet_batch_is_empty(batch)) {
break;
}
- pcap_batch_execute_conntrack(&ct, &pkt_batch);
+ pcap_batch_execute_conntrack(&ct, batch);
- for (i = 0; i < pkt_batch.count; i++) {
+ DP_PACKET_BATCH_FOR_EACH (packet, batch) {
struct ds ds = DS_EMPTY_INITIALIZER;
- struct dp_packet *pkt = pkt_batch.packets[i];
total_count++;
- format_flags(&ds, ct_state_to_string, pkt->md.ct_state, '|');
+ format_flags(&ds, ct_state_to_string, packet->md.ct_state, '|');
printf("%"PRIuSIZE": %s\n", total_count, ds_cstr(&ds));
ds_destroy(&ds);
}
- dp_packet_delete_batch(&pkt_batch, true);
- if (err) {
- break;
- }
+ dp_packet_delete_batch(batch, true);
}
conntrack_destroy(&ct);
}
--
1.9.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [userspace meter v3 3/5] dpif: Meter framework.
2017-01-31 6:39 [userspace meter v3 0/5] Userspace meter Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 1/5] netdev-dummy: Add --len option for netdev-dummy/receive command Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 2/5] dp-packet: Enhance packet batch APIs Andy Zhou
@ 2017-01-31 6:39 ` Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 4/5] ofproto: Meter translation Andy Zhou
` (2 subsequent siblings)
5 siblings, 0 replies; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 6:39 UTC (permalink / raw)
To: davem; +Cc: netdev, Jarno Rajahalme, Andy Zhou
From: Jarno Rajahalme <jarno@ovn.org>
Add DPIF-level infrastructure for meters. Allow meter_set to modify
the meter configuration (e.g. set the burst size if unspecified).
Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Signed-off-by: Andy Zhou <azhou@ovn.org>
---
datapath/linux/compat/include/linux/openvswitch.h | 4 +-
lib/dpif-netdev.c | 45 ++++++++++++
lib/dpif-netlink.c | 46 +++++++++++-
lib/dpif-provider.h | 29 ++++++++
lib/dpif.c | 88 +++++++++++++++++++++++
lib/dpif.h | 13 +++-
lib/odp-execute.c | 3 +
lib/odp-util.c | 14 ++++
ofproto/ofproto-dpif-sflow.c | 1 +
ofproto/ofproto-dpif.c | 60 ++++++++++++++--
ofproto/ofproto-provider.h | 13 ++--
ofproto/ofproto.c | 2 +-
12 files changed, 304 insertions(+), 14 deletions(-)
diff --git a/datapath/linux/compat/include/linux/openvswitch.h b/datapath/linux/compat/include/linux/openvswitch.h
index 425d3a4..b121391 100644
--- a/datapath/linux/compat/include/linux/openvswitch.h
+++ b/datapath/linux/compat/include/linux/openvswitch.h
@@ -787,13 +787,14 @@ enum ovs_nat_attr {
* fields within a header are modifiable, e.g. the IPv4 protocol and fragment
* type may not be changed.
*
- *
* @OVS_ACTION_ATTR_SET_TO_MASKED: Kernel internal masked set action translated
* from the @OVS_ACTION_ATTR_SET.
* @OVS_ACTION_ATTR_TUNNEL_PUSH: Push tunnel header described by struct
* ovs_action_push_tnl.
* @OVS_ACTION_ATTR_TUNNEL_POP: Lookup tunnel port by port-no passed and pop
* tunnel header.
+ * @OVS_ACTION_ATTR_METER: Run packet through a meter, which may drop the
+ * packet, or modify the packet (e.g., change the DSCP field).
*/
enum ovs_action_attr {
@@ -819,6 +820,7 @@ enum ovs_action_attr {
OVS_ACTION_ATTR_TUNNEL_PUSH, /* struct ovs_action_push_tnl*/
OVS_ACTION_ATTR_TUNNEL_POP, /* u32 port number. */
OVS_ACTION_ATTR_CLONE, /* Nested OVS_CLONE_ATTR_*. */
+ OVS_ACTION_ATTR_METER, /* u32 meter number. */
#endif
__OVS_ACTION_ATTR_MAX, /* Nothing past this will be accepted
* from userspace. */
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index 719a518..f81b464 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -3597,6 +3597,46 @@ dp_netdev_disable_upcall(struct dp_netdev *dp)
fat_rwlock_wrlock(&dp->upcall_rwlock);
}
+\f
+/* Meters */
+static void
+dpif_netdev_meter_get_features(const struct dpif * dpif OVS_UNUSED,
+ struct ofputil_meter_features *features)
+{
+ features->max_meters = 0;
+ features->band_types = 0;
+ features->capabilities = 0;
+ features->max_bands = 0;
+ features->max_color = 0;
+}
+
+static int
+dpif_netdev_meter_set(struct dpif *dpif OVS_UNUSED,
+ ofproto_meter_id *meter_id OVS_UNUSED,
+ struct ofputil_meter_config *config OVS_UNUSED)
+{
+ return EFBIG; /* meter_id out of range */
+}
+
+static int
+dpif_netdev_meter_get(const struct dpif *dpif OVS_UNUSED,
+ ofproto_meter_id meter_id OVS_UNUSED,
+ struct ofputil_meter_stats *stats OVS_UNUSED,
+ uint16_t n_bands OVS_UNUSED)
+{
+ return EFBIG; /* meter_id out of range */
+}
+
+static int
+dpif_netdev_meter_del(struct dpif *dpif OVS_UNUSED,
+ ofproto_meter_id meter_id OVS_UNUSED,
+ struct ofputil_meter_stats *stats OVS_UNUSED,
+ uint16_t n_bands OVS_UNUSED)
+{
+ return EFBIG; /* meter_id out of range */
+}
+
+\f
static void
dpif_netdev_disable_upcall(struct dpif *dpif)
OVS_NO_THREAD_SAFETY_ANALYSIS
@@ -4721,6 +4761,7 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
break;
}
+ case OVS_ACTION_ATTR_METER:
case OVS_ACTION_ATTR_PUSH_VLAN:
case OVS_ACTION_ATTR_POP_VLAN:
case OVS_ACTION_ATTR_PUSH_MPLS:
@@ -4858,6 +4899,10 @@ const struct dpif_class dpif_netdev_class = {
dpif_netdev_ct_dump_next,
dpif_netdev_ct_dump_done,
dpif_netdev_ct_flush,
+ dpif_netdev_meter_get_features,
+ dpif_netdev_meter_set,
+ dpif_netdev_meter_get,
+ dpif_netdev_meter_del,
};
static void
diff --git a/lib/dpif-netlink.c b/lib/dpif-netlink.c
index c8b0e37..8a48227 100644
--- a/lib/dpif-netlink.c
+++ b/lib/dpif-netlink.c
@@ -2356,6 +2356,46 @@ dpif_netlink_ct_flush(struct dpif *dpif OVS_UNUSED, const uint16_t *zone)
}
}
+\f
+/* Meters */
+static void
+dpif_netlink_meter_get_features(const struct dpif * dpif OVS_UNUSED,
+ struct ofputil_meter_features *features)
+{
+ features->max_meters = 0;
+ features->band_types = 0;
+ features->capabilities = 0;
+ features->max_bands = 0;
+ features->max_color = 0;
+}
+
+static int
+dpif_netlink_meter_set(struct dpif *dpif OVS_UNUSED,
+ ofproto_meter_id *meter_id OVS_UNUSED,
+ struct ofputil_meter_config *config OVS_UNUSED)
+{
+ return EFBIG; /* meter_id out of range */
+}
+
+static int
+dpif_netlink_meter_get(const struct dpif *dpif OVS_UNUSED,
+ ofproto_meter_id meter_id OVS_UNUSED,
+ struct ofputil_meter_stats *stats OVS_UNUSED,
+ uint16_t n_bands OVS_UNUSED)
+{
+ return EFBIG; /* meter_id out of range */
+}
+
+static int
+dpif_netlink_meter_del(struct dpif *dpif OVS_UNUSED,
+ ofproto_meter_id meter_id OVS_UNUSED,
+ struct ofputil_meter_stats *stats OVS_UNUSED,
+ uint16_t n_bands OVS_UNUSED)
+{
+ return EFBIG; /* meter_id out of range */
+}
+
+\f
const struct dpif_class dpif_netlink_class = {
"system",
NULL, /* init */
@@ -2400,7 +2440,11 @@ const struct dpif_class dpif_netlink_class = {
dpif_netlink_ct_dump_start,
dpif_netlink_ct_dump_next,
dpif_netlink_ct_dump_done,
- dpif_netlink_ct_flush
+ dpif_netlink_ct_flush,
+ dpif_netlink_meter_get_features,
+ dpif_netlink_meter_set,
+ dpif_netlink_meter_get,
+ dpif_netlink_meter_del,
};
static int
diff --git a/lib/dpif-provider.h b/lib/dpif-provider.h
index d3b2bb9..ea3e152 100644
--- a/lib/dpif-provider.h
+++ b/lib/dpif-provider.h
@@ -427,6 +427,35 @@ struct dpif_class {
/* Flushes the connection tracking tables. If 'zone' is not NULL,
* only deletes connections in '*zone'. */
int (*ct_flush)(struct dpif *, const uint16_t *zone);
+
+ /* Meters */
+
+ /* Queries 'dpif' for supported meter features.
+ * NULL pointer means no meter features are supported. */
+ void (*meter_get_features)(const struct dpif *,
+ struct ofputil_meter_features *);
+
+ /* Adds or modifies 'meter' in 'dpif'. If '*meter_id' is UINT32_MAX,
+ * adds a new meter, otherwise modifies an existing meter.
+ *
+ * If meter is successfully added, sets '*meter_id' to the new meter's
+ * meter id selected by 'dpif'. */
+ int (*meter_set)(struct dpif *, ofproto_meter_id *meter_id,
+ struct ofputil_meter_config *);
+
+ /* Queries 'dpif' for meter stats with the given 'meter_id'. Stores
+ * maximum of 'n_bands' meter statistics, returning the number of band
+ * stats returned in 'stats->n_bands' if successful. */
+ int (*meter_get)(const struct dpif *, ofproto_meter_id meter_id,
+ struct ofputil_meter_stats *, uint16_t n_bands);
+
+ /* Removes meter 'meter_id' from 'dpif'. Stores meter and band statistics
+ * (for maximum of 'n_bands', returning the number of band stats returned
+ * in 'stats->n_bands' if successful. 'stats' may be passed in as NULL if
+ * no stats are needed, in which case 'n_bands' must be passed in as
+ * zero. */
+ int (*meter_del)(struct dpif *, ofproto_meter_id meter_id,
+ struct ofputil_meter_stats *, uint16_t n_bands);
};
extern const struct dpif_class dpif_netlink_class;
diff --git a/lib/dpif.c b/lib/dpif.c
index 374f013..4e9476c 100644
--- a/lib/dpif.c
+++ b/lib/dpif.c
@@ -63,6 +63,9 @@ COVERAGE_DEFINE(dpif_flow_del);
COVERAGE_DEFINE(dpif_execute);
COVERAGE_DEFINE(dpif_purge);
COVERAGE_DEFINE(dpif_execute_with_help);
+COVERAGE_DEFINE(dpif_meter_set);
+COVERAGE_DEFINE(dpif_meter_get);
+COVERAGE_DEFINE(dpif_meter_del);
static const struct dpif_class *base_dpif_classes[] = {
#if defined(__linux__) || defined(_WIN32)
@@ -1174,6 +1177,7 @@ dpif_execute_helper_cb(void *aux_, struct dp_packet_batch *packets_,
}
case OVS_ACTION_ATTR_HASH:
+ case OVS_ACTION_ATTR_METER:
case OVS_ACTION_ATTR_PUSH_VLAN:
case OVS_ACTION_ATTR_POP_VLAN:
case OVS_ACTION_ATTR_PUSH_MPLS:
@@ -1757,3 +1761,87 @@ dpif_supports_tnl_push_pop(const struct dpif *dpif)
{
return dpif_is_netdev(dpif);
}
+
+/* Meters */
+void
+dpif_meter_get_features(const struct dpif *dpif,
+ struct ofputil_meter_features *features)
+{
+ memset(features, 0, sizeof *features);
+ if (dpif->dpif_class->meter_get_features) {
+ dpif->dpif_class->meter_get_features(dpif, features);
+ }
+}
+
+/* Adds or modifies meter identified by 'meter_id' in 'dpif'. If '*meter_id'
+ * is UINT32_MAX, adds a new meter, otherwise modifies an existing meter.
+ *
+ * If meter is successfully added, sets '*meter_id' to the new meter's
+ * meter number. */
+int
+dpif_meter_set(struct dpif *dpif, ofproto_meter_id *meter_id,
+ struct ofputil_meter_config *config)
+{
+ int error;
+
+ COVERAGE_INC(dpif_meter_set);
+
+ error = dpif->dpif_class->meter_set(dpif, meter_id, config);
+ if (!error) {
+ VLOG_DBG_RL(&dpmsg_rl, "%s: DPIF meter %"PRIu32" set",
+ dpif_name(dpif), meter_id->uint32);
+ } else {
+ VLOG_WARN_RL(&error_rl, "%s: failed to set DPIF meter %"PRIu32": %s",
+ dpif_name(dpif), meter_id->uint32, ovs_strerror(error));
+ meter_id->uint32 = UINT32_MAX;
+ }
+ return error;
+}
+
+int
+dpif_meter_get(const struct dpif *dpif, ofproto_meter_id meter_id,
+ struct ofputil_meter_stats *stats, uint16_t n_bands)
+{
+ int error;
+
+ COVERAGE_INC(dpif_meter_get);
+
+ error = dpif->dpif_class->meter_get(dpif, meter_id, stats, n_bands);
+ if (!error) {
+ VLOG_DBG_RL(&dpmsg_rl, "%s: DPIF meter %"PRIu32" get stats",
+ dpif_name(dpif), meter_id.uint32);
+ } else {
+ VLOG_WARN_RL(&error_rl,
+ "%s: failed to get DPIF meter %"PRIu32" stats: %s",
+ dpif_name(dpif), meter_id.uint32, ovs_strerror(error));
+ stats->packet_in_count = ~0;
+ stats->byte_in_count = ~0;
+ stats->n_bands = 0;
+ }
+ return error;
+}
+
+int
+dpif_meter_del(struct dpif *dpif, ofproto_meter_id meter_id,
+ struct ofputil_meter_stats *stats, uint16_t n_bands)
+{
+ int error;
+
+ COVERAGE_INC(dpif_meter_del);
+
+ error = dpif->dpif_class->meter_del(dpif, meter_id, stats, n_bands);
+ if (!error) {
+ VLOG_DBG_RL(&dpmsg_rl, "%s: DPIF meter %"PRIu32" deleted",
+ dpif_name(dpif), meter_id.uint32);
+ } else {
+ VLOG_WARN_RL(&error_rl,
+ "%s: failed to delete DPIF meter %"PRIu32": %s",
+ dpif_name(dpif), meter_id.uint32, ovs_strerror(error));
+ if (stats) {
+ stats->packet_in_count = ~0;
+ stats->byte_in_count = ~0;
+ stats->n_bands = 0;
+ }
+ }
+ return error;
+}
diff --git a/lib/dpif.h b/lib/dpif.h
index aa4fb8b..851fdc2 100644
--- a/lib/dpif.h
+++ b/lib/dpif.h
@@ -389,9 +389,10 @@
#include <stdint.h>
#include "dpdk.h"
-#include "netdev.h"
#include "dp-packet.h"
+#include "netdev.h"
#include "openflow/openflow.h"
+#include "openvswitch/ofp-util.h"
#include "ovs-numa.h"
#include "packets.h"
#include "util.h"
@@ -856,6 +857,16 @@ void dpif_disable_upcall(struct dpif *);
void dpif_print_packet(struct dpif *, struct dpif_upcall *);
\f
+/* Meters. */
+void dpif_meter_get_features(const struct dpif *,
+ struct ofputil_meter_features *);
+int dpif_meter_set(struct dpif *, ofproto_meter_id *meter_id,
+ struct ofputil_meter_config *);
+int dpif_meter_get(const struct dpif *, ofproto_meter_id meter_id,
+ struct ofputil_meter_stats *, uint16_t n_bands);
+int dpif_meter_del(struct dpif *, ofproto_meter_id meter_id,
+ struct ofputil_meter_stats *, uint16_t n_bands);
+\f
/* Miscellaneous. */
void dpif_get_netflow_ids(const struct dpif *,
diff --git a/lib/odp-execute.c b/lib/odp-execute.c
index 465280b..dd31c98 100644
--- a/lib/odp-execute.c
+++ b/lib/odp-execute.c
@@ -561,6 +561,7 @@ requires_datapath_assistance(const struct nlattr *a)
case OVS_ACTION_ATTR_USERSPACE:
case OVS_ACTION_ATTR_RECIRC:
case OVS_ACTION_ATTR_CT:
+ case OVS_ACTION_ATTR_METER:
return true;
case OVS_ACTION_ATTR_SET:
@@ -714,6 +715,8 @@ odp_execute_actions(void *dp, struct dp_packet_batch *batch, bool steal,
* stolen them. */
return;
}
+ case OVS_ACTION_ATTR_METER:
+ /* Not implemented yet. */
break;
case OVS_ACTION_ATTR_OUTPUT:
diff --git a/lib/odp-util.c b/lib/odp-util.c
index 430793b..5545f75 100644
--- a/lib/odp-util.c
+++ b/lib/odp-util.c
@@ -110,6 +110,7 @@ odp_action_len(uint16_t type)
case OVS_ACTION_ATTR_TRUNC: return sizeof(struct ovs_action_trunc);
case OVS_ACTION_ATTR_TUNNEL_PUSH: return ATTR_LEN_VARIABLE;
case OVS_ACTION_ATTR_TUNNEL_POP: return sizeof(uint32_t);
+ case OVS_ACTION_ATTR_METER: return sizeof(uint32_t);
case OVS_ACTION_ATTR_USERSPACE: return ATTR_LEN_VARIABLE;
case OVS_ACTION_ATTR_PUSH_VLAN: return sizeof(struct ovs_action_push_vlan);
case OVS_ACTION_ATTR_POP_VLAN: return 0;
@@ -794,6 +795,9 @@ format_odp_action(struct ds *ds, const struct nlattr *a)
}
switch (type) {
+ case OVS_ACTION_ATTR_METER:
+ ds_put_format(ds, "meter(%"PRIu32")", nl_attr_get_u32(a));
+ break;
case OVS_ACTION_ATTR_OUTPUT:
ds_put_format(ds, "%"PRIu32, nl_attr_get_u32(a));
break;
@@ -1683,6 +1687,16 @@ parse_odp_action(const char *s, const struct simap *port_names,
}
{
+ unsigned long long int meter_id;
+ int n = -1;
+
+ if (sscanf(s, "meter(%lli)%n", &meter_id, &n) > 0 && n > 0) {
+ nl_msg_put_u32(actions, OVS_ACTION_ATTR_METER, meter_id);
+ return n;
+ }
+ }
+
+ {
double percentage;
int n = -1;
diff --git a/ofproto/ofproto-dpif-sflow.c b/ofproto/ofproto-dpif-sflow.c
index e4ae760..f10ccd0 100644
--- a/ofproto/ofproto-dpif-sflow.c
+++ b/ofproto/ofproto-dpif-sflow.c
@@ -1132,6 +1132,7 @@ dpif_sflow_read_actions(const struct flow *flow,
case OVS_ACTION_ATTR_RECIRC:
case OVS_ACTION_ATTR_HASH:
case OVS_ACTION_ATTR_CT:
+ case OVS_ACTION_ATTR_METER:
break;
case OVS_ACTION_ATTR_SET_MASKED:
diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c
index e1112eb..cf9fb21 100644
--- a/ofproto/ofproto-dpif.c
+++ b/ofproto/ofproto-dpif.c
@@ -5185,6 +5185,58 @@ ofproto_dpif_delete_internal_flow(struct ofproto_dpif *ofproto,
return 0;
}
+static void
+meter_get_features(const struct ofproto *ofproto_,
+ struct ofputil_meter_features *features)
+{
+ const struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ dpif_meter_get_features(ofproto->backer->dpif, features);
+}
+
+static enum ofperr
+meter_set(struct ofproto *ofproto_, ofproto_meter_id *meter_id,
+ struct ofputil_meter_config *config)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ switch (dpif_meter_set(ofproto->backer->dpif, meter_id, config)) {
+ case 0:
+ return 0;
+ case EFBIG: /* meter_id out of range */
+ case ENOMEM: /* Cannot allocate meter */
+ return OFPERR_OFPMMFC_OUT_OF_METERS;
+ case EBADF: /* Unsupported flags */
+ return OFPERR_OFPMMFC_BAD_FLAGS;
+ case EINVAL: /* Too many bands */
+ return OFPERR_OFPMMFC_OUT_OF_BANDS;
+ case ENODEV: /* Unsupported band type */
+ return OFPERR_OFPMMFC_BAD_BAND;
+ default:
+ return OFPERR_OFPMMFC_UNKNOWN;
+ }
+}
+
+static enum ofperr
+meter_get(const struct ofproto *ofproto_, ofproto_meter_id meter_id,
+ struct ofputil_meter_stats *stats, uint16_t n_bands)
+{
+ const struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ if (!dpif_meter_get(ofproto->backer->dpif, meter_id, stats, n_bands)) {
+ return 0;
+ }
+ return OFPERR_OFPMMFC_UNKNOWN_METER;
+}
+
+static void
+meter_del(struct ofproto *ofproto_, ofproto_meter_id meter_id)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ dpif_meter_del(ofproto->backer->dpif, meter_id, NULL, 0);
+}
+
const struct ofproto_class ofproto_dpif_class = {
init,
enumerate_types,
@@ -5273,10 +5325,10 @@ const struct ofproto_class ofproto_dpif_class = {
set_mac_table_config,
set_mcast_snooping,
set_mcast_snooping_port,
- NULL, /* meter_get_features */
- NULL, /* meter_set */
- NULL, /* meter_get */
- NULL, /* meter_del */
+ meter_get_features,
+ meter_set,
+ meter_get,
+ meter_del,
group_alloc, /* group_alloc */
group_construct, /* group_construct */
group_destruct, /* group_destruct */
diff --git a/ofproto/ofproto-provider.h b/ofproto/ofproto-provider.h
index 3739ebc..abe9cb8 100644
--- a/ofproto/ofproto-provider.h
+++ b/ofproto/ofproto-provider.h
@@ -1769,16 +1769,17 @@ struct ofproto_class {
* leaving '*id' unchanged. On failure, the existing meter configuration
* is left intact. */
enum ofperr (*meter_set)(struct ofproto *ofproto, ofproto_meter_id *id,
- const struct ofputil_meter_config *config);
+ struct ofputil_meter_config *config);
/* Gets the meter and meter band packet and byte counts for maximum of
- * 'stats->n_bands' bands for the meter with provider ID 'id' within
- * 'ofproto'. The caller fills in the other stats values. The band stats
- * are copied to memory at 'stats->bands' provided by the caller. The
- * number of returned band stats is returned in 'stats->n_bands'. */
+ * 'n_bands' bands for the meter with provider ID 'id' within 'ofproto'.
+ * The caller fills in the other stats values. The band stats are copied
+ * to memory at 'stats->bands' provided by the caller. The number of
+ * returned band stats is returned in 'stats->n_bands'. */
enum ofperr (*meter_get)(const struct ofproto *ofproto,
ofproto_meter_id id,
- struct ofputil_meter_stats *stats);
+ struct ofputil_meter_stats *stats,
+ uint16_t n_bands);
/* Deletes a meter, making the 'ofproto_meter_id' invalid for any
* further calls. */
diff --git a/ofproto/ofproto.c b/ofproto/ofproto.c
index 0b5e0fa..49652d7 100644
--- a/ofproto/ofproto.c
+++ b/ofproto/ofproto.c
@@ -6423,7 +6423,7 @@ handle_meter_request(struct ofconn *ofconn, const struct ofp_header *request,
if (!ofproto->ofproto_class->meter_get(ofproto,
meter->provider_meter_id,
- &stats)) {
+ &stats, meter->n_bands)) {
ofputil_append_meter_stats(&replies, &stats);
}
} else { /* type == OFPTYPE_METER_CONFIG_REQUEST */
--
1.9.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [userspace meter v3 4/5] ofproto: Meter translation.
2017-01-31 6:39 [userspace meter v3 0/5] Userspace meter Andy Zhou
` (2 preceding siblings ...)
2017-01-31 6:39 ` [userspace meter v3 3/5] dpif: Meter framework Andy Zhou
@ 2017-01-31 6:39 ` Andy Zhou
2017-01-31 6:40 ` [userspace meter v3 5/5] dpif-netdev: Simple DROP meter implementation Andy Zhou
2017-01-31 7:05 ` [userspace meter v3 0/5] Userspace meter Andy Zhou
5 siblings, 0 replies; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 6:39 UTC (permalink / raw)
To: davem; +Cc: netdev, Jarno Rajahalme, Andy Zhou
From: Jarno Rajahalme <jarno@ovn.org>
Translate OpenFlow METER instructions to datapath meter actions.
Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Signed-off-by: Andy Zhou <azhou@ovn.org>
---
include/openvswitch/ofp-actions.h | 1 +
lib/dpif.c | 40 +++++++++++++++++++++++++-------
lib/ofp-actions.c | 1 +
ofproto/ofproto-dpif-xlate.c | 11 ++++++++-
ofproto/ofproto.c | 49 ++++++++++++++++++++++-----------------
5 files changed, 72 insertions(+), 30 deletions(-)
diff --git a/include/openvswitch/ofp-actions.h b/include/openvswitch/ofp-actions.h
index 8ca787a..e269901 100644
--- a/include/openvswitch/ofp-actions.h
+++ b/include/openvswitch/ofp-actions.h
@@ -532,6 +532,7 @@ struct ofpact_metadata {
struct ofpact_meter {
struct ofpact ofpact;
uint32_t meter_id;
+ uint32_t provider_meter_id;
};
/* OFPACT_WRITE_ACTIONS, OFPACT_CLONE.
diff --git a/lib/dpif.c b/lib/dpif.c
index 4e9476c..cc49d94 100644
--- a/lib/dpif.c
+++ b/lib/dpif.c
@@ -1104,6 +1104,7 @@ struct dpif_execute_helper_aux {
struct dpif *dpif;
const struct flow *flow;
int error;
+ const struct nlattr *meter_action; /* Non-NULL, if have a meter action. */
};
/* This is called for actions that need the context of the datapath to be
@@ -1119,6 +1120,13 @@ dpif_execute_helper_cb(void *aux_, struct dp_packet_batch *packets_,
ovs_assert(packets_->count == 1);
switch ((enum ovs_action_attr)type) {
+ case OVS_ACTION_ATTR_METER:
+ /* Maintain a pointer to the first meter action seen. */
+ if (!aux->meter_action) {
+ aux->meter_action = action;
+ }
+ break;
+
case OVS_ACTION_ATTR_CT:
case OVS_ACTION_ATTR_OUTPUT:
case OVS_ACTION_ATTR_TUNNEL_PUSH:
@@ -1129,15 +1137,29 @@ dpif_execute_helper_cb(void *aux_, struct dp_packet_batch *packets_,
struct ofpbuf execute_actions;
uint64_t stub[256 / 8];
struct pkt_metadata *md = &packet->md;
- bool dst_set;
- dst_set = flow_tnl_dst_is_set(&md->tunnel);
- if (dst_set) {
+ if (flow_tnl_dst_is_set(&md->tunnel) || aux->meter_action) {
+ ofpbuf_use_stub(&execute_actions, stub, sizeof stub);
+
+ if (aux->meter_action) {
+ const struct nlattr *a = aux->meter_action;
+
+ do {
+ ofpbuf_put(&execute_actions, a, NLA_ALIGN(a->nla_len));
+ /* Find next meter action before 'action', if any. */
+ do {
+ a = nl_attr_next(a);
+ } while (a != action &&
+ nl_attr_type(a) != OVS_ACTION_ATTR_METER);
+ } while (a != action);
+ }
+
/* The Linux kernel datapath throws away the tunnel information
* that we supply as metadata. We have to use a "set" action to
* supply it. */
- ofpbuf_use_stub(&execute_actions, stub, sizeof stub);
- odp_put_tunnel_action(&md->tunnel, &execute_actions);
+ if (md->tunnel.ip_dst) {
+ odp_put_tunnel_action(&md->tunnel, &execute_actions);
+ }
ofpbuf_put(&execute_actions, action, NLA_ALIGN(action->nla_len));
execute.actions = execute_actions.data;
@@ -1170,14 +1192,16 @@ dpif_execute_helper_cb(void *aux_, struct dp_packet_batch *packets_,
dp_packet_delete(clone);
- if (dst_set) {
+ if (flow_tnl_dst_is_set(&md->tunnel) || aux->meter_action) {
ofpbuf_uninit(&execute_actions);
+
+ /* Do not re-use the same meters for later output actions. */
+ aux->meter_action = NULL;
}
break;
}
case OVS_ACTION_ATTR_HASH:
- case OVS_ACTION_ATTR_METER:
case OVS_ACTION_ATTR_PUSH_VLAN:
case OVS_ACTION_ATTR_POP_VLAN:
case OVS_ACTION_ATTR_PUSH_MPLS:
@@ -1201,7 +1225,7 @@ dpif_execute_helper_cb(void *aux_, struct dp_packet_batch *packets_,
static int
dpif_execute_with_help(struct dpif *dpif, struct dpif_execute *execute)
{
- struct dpif_execute_helper_aux aux = {dpif, execute->flow, 0};
+ struct dpif_execute_helper_aux aux = {dpif, execute->flow, 0, NULL};
struct dp_packet_batch pb;
COVERAGE_INC(dpif_execute_with_help);
diff --git a/lib/ofp-actions.c b/lib/ofp-actions.c
index cf1ad0f..733b2c5 100644
--- a/lib/ofp-actions.c
+++ b/lib/ofp-actions.c
@@ -6868,6 +6868,7 @@ ofpacts_pull_openflow_instructions(struct ofpbuf *openflow,
om = ofpact_put_METER(ofpacts);
om->meter_id = ntohl(oim->meter_id);
+ om->provider_meter_id = UINT32_MAX; /* No provider meter ID. */
}
if (insts[OVSINST_OFPIT11_APPLY_ACTIONS]) {
const struct ofp_action_header *actions;
diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c
index 48b27a6..166e236 100644
--- a/ofproto/ofproto-dpif-xlate.c
+++ b/ofproto/ofproto-dpif-xlate.c
@@ -4579,6 +4579,15 @@ xlate_clone(struct xlate_ctx *ctx, const struct ofpact_nest *oc)
ctx->was_mpls = old_was_mpls;
}
+static void
+xlate_meter_action(struct xlate_ctx *ctx, const struct ofpact_meter *meter)
+{
+ if (meter->provider_meter_id != UINT32_MAX) {
+ nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_METER,
+ meter->provider_meter_id);
+ }
+}
+
static bool
may_receive(const struct xport *xport, struct xlate_ctx *ctx)
{
@@ -5388,7 +5397,7 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len,
break;
case OFPACT_METER:
- /* Not implemented yet. */
+ xlate_meter_action(ctx, ofpact_get_METER(a));
break;
case OFPACT_GOTO_TABLE: {
diff --git a/ofproto/ofproto.c b/ofproto/ofproto.c
index 49652d7..f5fa897 100644
--- a/ofproto/ofproto.c
+++ b/ofproto/ofproto.c
@@ -2989,8 +2989,8 @@ remove_groups_rcu(struct ofgroup **groups)
free(groups);
}
-static uint32_t get_provider_meter_id(const struct ofproto *,
- uint32_t of_meter_id);
+static bool ofproto_fix_meter_action(const struct ofproto *,
+ struct ofpact_meter *);
/* Creates and returns a new 'struct rule_actions', whose actions are a copy
* of from the 'ofpacts_len' bytes of 'ofpacts'. */
@@ -3383,6 +3383,7 @@ reject_slave_controller(struct ofconn *ofconn)
* for 'ofproto':
*
* - If they use a meter, then 'ofproto' has that meter configured.
+ * Updates the meter action with ofproto's datapath's provider_meter_id.
*
* - If they use any groups, then 'ofproto' has that group configured.
*
@@ -3392,18 +3393,17 @@ reject_slave_controller(struct ofconn *ofconn)
enum ofperr
ofproto_check_ofpacts(struct ofproto *ofproto,
const struct ofpact ofpacts[], size_t ofpacts_len)
- OVS_REQUIRES(ofproto_mutex)
{
- uint32_t mid;
+ const struct ofpact *a;
- mid = ofpacts_get_meter(ofpacts, ofpacts_len);
- if (mid && get_provider_meter_id(ofproto, mid) == UINT32_MAX) {
- return OFPERR_OFPMMFC_INVALID_METER;
- }
+ OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) {
+ if (a->type == OFPACT_METER &&
+ !ofproto_fix_meter_action(ofproto, ofpact_get_METER(a))) {
+ return OFPERR_OFPMMFC_INVALID_METER;
+ }
- const struct ofpact_group *a;
- OFPACT_FOR_EACH_TYPE_FLATTENED (a, GROUP, ofpacts, ofpacts_len) {
- if (!ofproto_group_exists(ofproto, a->group_id)) {
+ if (a->type == OFPACT_GROUP
+ && !ofproto_group_exists(ofproto, ofpact_get_GROUP(a)->group_id)) {
return OFPERR_OFPBAC_BAD_OUT_GROUP;
}
}
@@ -6133,20 +6133,27 @@ struct meter {
};
/*
- * This is used in instruction validation at flow set-up time,
- * as flows may not use non-existing meters.
- * Return value of UINT32_MAX signifies an invalid meter.
+ * This is used in instruction validation at flow set-up time, to map
+ * the OpenFlow meter ID to the corresponding datapath provider meter
+ * ID. If either does not exist, returns false. Otherwise updates
+ * the meter action and returns true.
*/
-static uint32_t
-get_provider_meter_id(const struct ofproto *ofproto, uint32_t of_meter_id)
+static bool
+ofproto_fix_meter_action(const struct ofproto *ofproto,
+ struct ofpact_meter *ma)
{
- if (of_meter_id && of_meter_id <= ofproto->meter_features.max_meters) {
- const struct meter *meter = ofproto->meters[of_meter_id];
- if (meter) {
- return meter->provider_meter_id.uint32;
+ if (ma->meter_id && ma->meter_id <= ofproto->meter_features.max_meters) {
+ const struct meter *meter = ofproto->meters[ma->meter_id];
+
+ if (meter && meter->provider_meter_id.uint32 != UINT32_MAX) {
+ /* Update the action with the provider's meter ID, so that we
+ * do not need any synchronization between ofproto_dpif_xlate
+ * and ofproto for meter table access. */
+ ma->provider_meter_id = meter->provider_meter_id.uint32;
+ return true;
}
}
- return UINT32_MAX;
+ return false;
}
/* Finds the meter invoked by 'rule''s actions and adds 'rule' to the meter's
--
1.9.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [userspace meter v3 5/5] dpif-netdev: Simple DROP meter implementation.
2017-01-31 6:39 [userspace meter v3 0/5] Userspace meter Andy Zhou
` (3 preceding siblings ...)
2017-01-31 6:39 ` [userspace meter v3 4/5] ofproto: Meter translation Andy Zhou
@ 2017-01-31 6:40 ` Andy Zhou
2017-01-31 7:05 ` [userspace meter v3 0/5] Userspace meter Andy Zhou
5 siblings, 0 replies; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 6:40 UTC (permalink / raw)
To: davem; +Cc: netdev, Jarno Rajahalme, Andy Zhou
From: Jarno Rajahalme <jarno@ovn.org>
Meters may be used by any flow, so some kind of locking must be used.
In this version we have an adaptive mutex for each meter, which may
not be optimal for DPDK. However, this should serve as a basis for
further improvement.
A batch of packets is first tried as a whole, and only if some of the
meter bands are hit, we need to process the packets individually.
Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Signed-off-by: Andy Zhou <azhou@ovn.org>
---
lib/dpif-netdev.c | 362 ++++++++++++++++++++++++++++++++++++++++++++++++---
tests/dpif-netdev.at | 106 +++++++++++++++
2 files changed, 450 insertions(+), 18 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index f81b464..96c283a 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -86,6 +86,8 @@ DEFINE_STATIC_PER_THREAD_DATA(uint32_t, recirc_depth, 0)
/* Configuration parameters. */
enum { MAX_FLOWS = 65536 }; /* Maximum number of flows in flow table. */
+enum { MAX_METERS = 65536 }; /* Maximum number of meters. */
+enum { MAX_BANDS = 8 }; /* Maximum number of bands / meter. */
/* Protects against changes to 'dp_netdevs'. */
static struct ovs_mutex dp_netdev_mutex = OVS_MUTEX_INITIALIZER;
@@ -193,6 +195,31 @@ static bool dpcls_lookup(struct dpcls *cls,
struct dpcls_rule **rules, size_t cnt,
int *num_lookups_p);
\f
+/* Set of supported meter flags */
+#define DP_SUPPORTED_METER_FLAGS_MASK \
+ (OFPMF13_STATS | OFPMF13_PKTPS | OFPMF13_KBPS | OFPMF13_BURST)
+
+/* Set of supported meter band types */
+#define DP_SUPPORTED_METER_BAND_TYPES \
+ ( 1 << OFPMBT13_DROP )
+
+struct dp_meter_band {
+ struct ofputil_meter_band up; /* type, prec_level, pad, rate, burst_size */
+ uint32_t bucket; /* In 1/1000 packets (for PKTPS), or in bits (for KBPS) */
+ uint64_t packet_count;
+ uint64_t byte_count;
+};
+
+struct dp_meter {
+ uint16_t flags;
+ uint16_t n_bands;
+ uint32_t max_delta_t;
+ uint64_t used;
+ uint64_t packet_count;
+ uint64_t byte_count;
+ struct dp_meter_band bands[];
+};
+
/* Datapath based on the network device interface from netdev.h.
*
*
@@ -223,6 +250,11 @@ struct dp_netdev {
struct hmap ports;
struct seq *port_seq; /* Incremented whenever a port changes. */
+ /* Meters. */
+ struct ovs_mutex meter_locks[MAX_METERS];
+ struct dp_meter *meters[MAX_METERS]; /* Meter bands. */
+ uint32_t meter_free; /* Next free meter. */
+
/* Protects access to ofproto-dpif-upcall interface during revalidator
* thread synchronization. */
struct fat_rwlock upcall_rwlock;
@@ -1059,6 +1091,10 @@ create_dp_netdev(const char *name, const struct dpif_class *class,
dp->reconfigure_seq = seq_create();
dp->last_reconfigure_seq = seq_read(dp->reconfigure_seq);
+ for (int i = 0; i < MAX_METERS; ++i) {
+ ovs_mutex_init_adaptive(&dp->meter_locks[i]);
+ }
+
/* Disable upcalls by default. */
dp_netdev_disable_upcall(dp);
dp->upcall_aux = NULL;
@@ -1136,6 +1172,16 @@ dp_netdev_destroy_upcall_lock(struct dp_netdev *dp)
fat_rwlock_destroy(&dp->upcall_rwlock);
}
+static void
+dp_delete_meter(struct dp_netdev *dp, uint32_t meter_id)
+ OVS_REQUIRES(dp->meter_locks[meter_id])
+{
+ if (dp->meters[meter_id]) {
+ free(dp->meters[meter_id]);
+ dp->meters[meter_id] = NULL;
+ }
+}
+
/* Requires dp_netdev_mutex so that we can't get a new reference to 'dp'
* through the 'dp_netdevs' shash while freeing 'dp'. */
static void
@@ -1151,6 +1197,7 @@ dp_netdev_free(struct dp_netdev *dp)
do_del_port(dp, port);
}
ovs_mutex_unlock(&dp->port_mutex);
+
dp_netdev_destroy_all_pmds(dp, true);
cmap_destroy(&dp->poll_threads);
@@ -1169,6 +1216,13 @@ dp_netdev_free(struct dp_netdev *dp)
/* Upcalls must be disabled at this point */
dp_netdev_destroy_upcall_lock(dp);
+ for (int i = 0; i < MAX_METERS; ++i) {
+ ovs_mutex_lock(&dp->meter_locks[i]);
+ dp_delete_meter(dp, i);
+ ovs_mutex_unlock(&dp->meter_locks[i]);
+ ovs_mutex_destroy(&dp->meter_locks[i]);
+ }
+
free(dp->pmd_cmask);
free(CONST_CAST(char *, dp->name));
free(dp);
@@ -3603,37 +3657,304 @@ static void
dpif_netdev_meter_get_features(const struct dpif * dpif OVS_UNUSED,
struct ofputil_meter_features *features)
{
- features->max_meters = 0;
- features->band_types = 0;
- features->capabilities = 0;
- features->max_bands = 0;
+ features->max_meters = MAX_METERS;
+ features->band_types = DP_SUPPORTED_METER_BAND_TYPES;
+ features->capabilities = DP_SUPPORTED_METER_FLAGS_MASK;
+ features->max_bands = MAX_BANDS;
features->max_color = 0;
}
+/* Returns false when packet needs to be dropped. */
+static void
+dp_netdev_run_meter(struct dp_netdev *dp, struct dp_packet_batch *packets_,
+ uint32_t meter_id, long long int now)
+{
+ struct dp_meter *meter;
+ struct dp_meter_band *band;
+ long long int long_delta_t; /* msec */
+ uint32_t delta_t; /* msec */
+ int i;
+ int cnt = packets_->count;
+ uint32_t bytes, volume;
+ int exceeded_band[NETDEV_MAX_BURST];
+ uint32_t exceeded_rate[NETDEV_MAX_BURST];
+ int exceeded_pkt = cnt; /* First packet that exceeded a band rate. */
+
+ if (meter_id >= MAX_METERS) {
+ return;
+ }
+
+ ovs_mutex_lock(&dp->meter_locks[meter_id]);
+ meter = dp->meters[meter_id];
+ if (!meter) {
+ goto out;
+ }
+
+ /* Initialize as negative values. */
+ memset(exceeded_band, 0xff, cnt * sizeof *exceeded_band);
+ /* Initialize as zeroes. */
+ memset(exceeded_rate, 0, cnt * sizeof *exceeded_rate);
+
+ /* All packets will hit the meter at the same time. */
+ long_delta_t = (now - meter->used); /* msec */
+
+ /* Make sure delta_t will not be too large, so that bucket will not
+ * wrap around below. */
+ delta_t = (long_delta_t > (long long int)meter->max_delta_t)
+ ? meter->max_delta_t : (uint32_t)long_delta_t;
+
+ /* Update meter stats. */
+ meter->used = now;
+ meter->packet_count += cnt;
+ bytes = 0;
+ for (i = 0; i < cnt; i++) {
+ bytes += dp_packet_size(packets_->packets[i]);
+ }
+ meter->byte_count += bytes;
+
+ /* Meters can operate in terms of packets per second or kilobits per
+ * second. */
+ if (meter->flags & OFPMF13_PKTPS) {
+ /* Rate in packets/second, bucket 1/1000 packets. */
+ /* msec * packets/sec = 1/1000 packets. */
+ volume = cnt * 1000; /* Take 'cnt' packets from the bucket. */
+ } else {
+ /* Rate in kbps, bucket in bits. */
+ /* msec * kbps = bits */
+ volume = bytes * 8;
+ }
+
+ /* Update all bands and find the one hit with the highest rate for each
+ * packet (if any). */
+ for (int m = 0; m < meter->n_bands; ++m) {
+ band = &meter->bands[m];
+
+ /* Update band's bucket. */
+ band->bucket += delta_t * band->up.rate;
+ if (band->bucket > band->up.burst_size) {
+ band->bucket = band->up.burst_size;
+ }
+
+ /* Drain the bucket for all the packets, if possible. */
+ if (band->bucket >= volume) {
+ band->bucket -= volume;
+ } else {
+ int band_exceeded_pkt;
+
+ /* Band limit hit, must process packet-by-packet. */
+ if (meter->flags & OFPMF13_PKTPS) {
+ band_exceeded_pkt = band->bucket / 1000;
+ band->bucket %= 1000; /* Remainder stays in bucket. */
+
+ /* Update the exceeding band for each exceeding packet.
+ * (Only one band will be fired by a packet, and that
+ * can be different for each packet.) */
+ for (i = band_exceeded_pkt; i < cnt; i++) {
+ if (band->up.rate > exceeded_rate[i]) {
+ exceeded_rate[i] = band->up.rate;
+ exceeded_band[i] = m;
+ }
+ }
+ } else {
+ /* Packet sizes differ, must process one-by-one. */
+ band_exceeded_pkt = cnt;
+ for (i = 0; i < cnt; i++) {
+ uint32_t bits = dp_packet_size(packets_->packets[i]) * 8;
+
+ if (band->bucket >= bits) {
+ band->bucket -= bits;
+ } else {
+ if (i < band_exceeded_pkt) {
+ band_exceeded_pkt = i;
+ }
+ /* Update the exceeding band for the exceeding packet.
+ * (Only one band will be fired by a packet, and that
+ * can be different for each packet.) */
+ if (band->up.rate > exceeded_rate[i]) {
+ exceeded_rate[i] = band->up.rate;
+ exceeded_band[i] = m;
+ }
+ }
+ }
+ }
+ /* Remember the first exceeding packet. */
+ if (exceeded_pkt > band_exceeded_pkt) {
+ exceeded_pkt = band_exceeded_pkt;
+ }
+ }
+ }
+
+ /* Fire the highest rate band exceeded by each packet.
+ * Drop packets if needed, by swapping packet to the end that will be
+ * ignored. */
+ const size_t size = dp_packet_batch_size(packets_);
+ struct dp_packet *packet;
+ size_t j;
+ DP_PACKET_BATCH_REFILL_FOR_EACH (j, size, packet, packets_) {
+ if (exceeded_band[j] >= 0) {
+ /* Meter drop packet. */
+ band = &meter->bands[exceeded_band[j]];
+ band->packet_count += 1;
+ band->byte_count += dp_packet_size(packet);
+
+ dp_packet_delete(packet);
+ } else {
+ /* Meter accepts packet. */
+ dp_packet_batch_refill(packets_, packet, j);
+ }
+ }
+ out:
+ ovs_mutex_unlock(&dp->meter_locks[meter_id]);
+}
+
+/* Meter set/get/del processing is still single-threaded. */
static int
-dpif_netdev_meter_set(struct dpif *dpif OVS_UNUSED,
- ofproto_meter_id *meter_id OVS_UNUSED,
- struct ofputil_meter_config *config OVS_UNUSED)
+dpif_netdev_meter_set(struct dpif *dpif, ofproto_meter_id *meter_id,
+ struct ofputil_meter_config *config)
{
- return EFBIG; /* meter_id out of range */
+ struct dp_netdev *dp = get_dp_netdev(dpif);
+ uint32_t mid = meter_id->uint32;
+ struct dp_meter *meter;
+ int i;
+
+ if (mid == UINT32_MAX) {
+ mid = dp->meter_free;
+ }
+ if (mid >= MAX_METERS) {
+ return EFBIG; /* Meter_id out of range. */
+ }
+
+ if (config->flags & ~DP_SUPPORTED_METER_FLAGS_MASK ||
+ !(config->flags & (OFPMF13_KBPS | OFPMF13_PKTPS))) {
+ return EBADF; /* Unsupported flags set */
+ }
+ /* Validate bands */
+ if (config->n_bands == 0 || config->n_bands > MAX_BANDS) {
+ return EINVAL; /* Too many bands */
+ }
+ for (i = 0; i < config->n_bands; ++i) {
+ switch (config->bands[i].type) {
+ case OFPMBT13_DROP:
+ break;
+ default:
+ return ENODEV; /* Unsupported band type */
+ }
+ }
+
+ /* Allocate meter */
+ meter = xzalloc(sizeof *meter
+ + config->n_bands * sizeof(struct dp_meter_band));
+ if (meter) {
+ meter->flags = config->flags;
+ meter->n_bands = config->n_bands;
+ meter->max_delta_t = 0;
+ meter->used = time_msec();
+
+ /* set up bands */
+ for (i = 0; i < config->n_bands; ++i) {
+ uint32_t band_max_delta_t;
+
+ /* Set burst size to a workable value if none specified. */
+ if (config->bands[i].burst_size == 0) {
+ config->bands[i].burst_size = config->bands[i].rate;
+ }
+
+ meter->bands[i].up = config->bands[i];
+ /* Convert burst size to the bucket units: */
+ /* pkts => 1/1000 packets, kilobits => bits. */
+ meter->bands[i].up.burst_size *= 1000;
+ /* Initialize bucket to empty. */
+ meter->bands[i].bucket = 0;
+
+ /* Figure out max delta_t that is enough to fill any bucket. */
+ band_max_delta_t
+ = meter->bands[i].up.burst_size / meter->bands[i].up.rate;
+ if (band_max_delta_t > meter->max_delta_t) {
+ meter->max_delta_t = band_max_delta_t;
+ }
+ }
+
+ ovs_mutex_lock(&dp->meter_locks[mid]);
+ dp_delete_meter(dp, mid); /* Free existing meter, if any */
+ dp->meters[mid] = meter;
+ ovs_mutex_unlock(&dp->meter_locks[mid]);
+
+ meter_id->uint32 = mid; /* Store on success. */
+
+ /* Find next free meter */
+ if (dp->meter_free == mid) { /* Now taken. */
+ do {
+ if (++mid >= MAX_METERS) { /* Wrap around */
+ mid = 0;
+ }
+ if (mid == dp->meter_free) { /* Full circle */
+ mid = MAX_METERS;
+ break;
+ }
+ } while (dp->meters[mid]);
+ dp->meter_free = mid; /* Next free meter or MAX_METERS */
+ }
+ return 0;
+ }
+ return ENOMEM;
}
static int
-dpif_netdev_meter_get(const struct dpif *dpif OVS_UNUSED,
- ofproto_meter_id meter_id OVS_UNUSED,
- struct ofputil_meter_stats *stats OVS_UNUSED,
- uint16_t n_bands OVS_UNUSED)
+dpif_netdev_meter_get(const struct dpif *dpif,
+ ofproto_meter_id meter_id_,
+ struct ofputil_meter_stats *stats, uint16_t n_bands)
{
- return EFBIG; /* meter_id out of range */
+ const struct dp_netdev *dp = get_dp_netdev(dpif);
+ const struct dp_meter *meter;
+ uint32_t meter_id = meter_id_.uint32;
+
+ if (meter_id >= MAX_METERS) {
+ return EFBIG;
+ }
+ meter = dp->meters[meter_id];
+ if (!meter) {
+ return ENOENT;
+ }
+ if (stats) {
+ int i = 0;
+
+ ovs_mutex_lock(&dp->meter_locks[meter_id]);
+ stats->packet_in_count = meter->packet_count;
+ stats->byte_in_count = meter->byte_count;
+
+ for (i = 0; i < n_bands && i < meter->n_bands; ++i) {
+ stats->bands[i].packet_count = meter->bands[i].packet_count;
+ stats->bands[i].byte_count = meter->bands[i].byte_count;
+ }
+ ovs_mutex_unlock(&dp->meter_locks[meter_id]);
+
+ stats->n_bands = i;
+ }
+ return 0;
}
static int
-dpif_netdev_meter_del(struct dpif *dpif OVS_UNUSED,
- ofproto_meter_id meter_id OVS_UNUSED,
- struct ofputil_meter_stats *stats OVS_UNUSED,
- uint16_t n_bands OVS_UNUSED)
+dpif_netdev_meter_del(struct dpif *dpif,
+ ofproto_meter_id meter_id_,
+ struct ofputil_meter_stats *stats, uint16_t n_bands)
{
- return EFBIG; /* meter_id out of range */
+ struct dp_netdev *dp = get_dp_netdev(dpif);
+ int error;
+
+ error = dpif_netdev_meter_get(dpif, meter_id_, stats, n_bands);
+ if (!error) {
+ uint32_t meter_id = meter_id_.uint32;
+
+ ovs_mutex_lock(&dp->meter_locks[meter_id]);
+ dp_delete_meter(dp, meter_id);
+ ovs_mutex_unlock(&dp->meter_locks[meter_id]);
+
+ /* Keep free meter index as low as possible */
+ if (meter_id < dp->meter_free) {
+ dp->meter_free = meter_id;
+ }
+ }
+ return error;
}
\f
@@ -4565,6 +4886,7 @@ dp_execute_userspace_action(struct dp_netdev_pmd_thread *pmd,
static void
dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
const struct nlattr *a, bool may_steal)
+ OVS_NO_THREAD_SAFETY_ANALYSIS
{
struct dp_netdev_execute_aux *aux = aux_;
uint32_t *depth = recirc_depth_get();
@@ -4762,6 +5084,10 @@ dp_execute_cb(void *aux_, struct dp_packet_batch *packets_,
}
case OVS_ACTION_ATTR_METER:
+ dp_netdev_run_meter(pmd->dp, packets_, nl_attr_get_u32(a),
+ time_msec());
+ break;
+
case OVS_ACTION_ATTR_PUSH_VLAN:
case OVS_ACTION_ATTR_POP_VLAN:
case OVS_ACTION_ATTR_PUSH_MPLS:
diff --git a/tests/dpif-netdev.at b/tests/dpif-netdev.at
index fff0460..586a5b1 100644
--- a/tests/dpif-netdev.at
+++ b/tests/dpif-netdev.at
@@ -4,6 +4,13 @@ m4_divert_push([PREPARE_TESTS])
[
# Strips out uninteresting parts of flow output, as well as parts
# that vary from one run to another (e.g., timing and bond actions).
+strip_timers () {
+ sed '
+ s/duration:[0-9]*\.[0-9]*/duration:0.0/
+ s/used:[0-9]*\.[0-9]*/used:0.0/
+'
+}
+
strip_xout () {
sed '
s/ufid:[-0-9a-f]* //
@@ -160,3 +167,102 @@ skb_priority(0/0),skb_mark(0/0),ct_state(0/0),ct_zone(0/0),ct_mark(0/0),ct_label
DPIF_NETDEV_MISS_FLOW_DUMP([dummy])
DPIF_NETDEV_MISS_FLOW_DUMP([dummy-pmd])
+
+AT_SETUP([dpif-netdev - meters])
+# Create br0 with interfaces p1 and p7
+# and br1 with interfaces p2 and p8
+# with p1 and p2 connected via unix domain socket
+OVS_VSWITCHD_START(
+ [add-port br0 p1 -- set interface p1 type=dummy options:pstream=punix:$OVS_RUNDIR/p0.sock ofport_request=1 -- \
+ add-port br0 p7 -- set interface p7 ofport_request=7 type=dummy -- \
+ add-br br1 -- \
+ set bridge br1 other-config:hwaddr=aa:66:aa:66:00:00 -- \
+ set bridge br1 datapath-type=dummy other-config:datapath-id=1234 \
+ fail-mode=secure -- \
+ add-port br1 p2 -- set interface p2 type=dummy options:stream=unix:$OVS_RUNDIR/p0.sock ofport_request=2 -- \
+ add-port br1 p8 -- set interface p8 ofport_request=8 type=dummy --])
+AT_CHECK([ovs-appctl vlog/set dpif:dbg dpif_netdev:dbg])
+
+AT_CHECK([ovs-ofctl -O OpenFlow13 add-meter br0 'meter=1 pktps burst stats bands=type=drop rate=1 burst_size=1'])
+AT_CHECK([ovs-ofctl -O OpenFlow13 add-meter br0 'meter=2 kbps burst stats bands=type=drop rate=1 burst_size=2'])
+AT_CHECK([ovs-ofctl -O OpenFlow13 add-flow br0 'in_port=1 action=meter:1,7'])
+AT_CHECK([ovs-ofctl -O OpenFlow13 add-flow br0 'in_port=7 action=meter:2,1'])
+AT_CHECK([ovs-ofctl add-flow br1 'in_port=2 action=8'])
+AT_CHECK([ovs-ofctl add-flow br1 'in_port=8 action=2'])
+ovs-appctl time/stop
+
+AT_CHECK([ovs-ofctl -O OpenFlow13 dump-meters br0], [0], [dnl
+OFPST_METER_CONFIG reply (OF1.3) (xid=0x2):
+meter=1 pktps burst stats bands=
+type=drop rate=1 burst_size=1
+
+meter=2 kbps burst stats bands=
+type=drop rate=1 burst_size=2
+])
+
+ovs-appctl time/warp 5000
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+sleep 1 # wait for forwarders process packets
+
+# Meter 1 is measuring packets, allowing one packet per second with
+# bursts of one packet, so 4 out of 5 packets should hit the drop
+# band.
+# Meter 2 is measuring kbps, with burst size 2 (== 2000 bits). 4 packets
+# (240 bytes == 1920 bits) pass, but the last packet should hit the drop band.
+AT_CHECK([ovs-ofctl -O OpenFlow13 meter-stats br0 | strip_timers], [0], [dnl
+OFPST_METER reply (OF1.3) (xid=0x2):
+meter:1 flow_count:1 packet_in_count:5 byte_in_count:300 duration:0.0s bands:
+0: packet_count:4 byte_count:240
+
+meter:2 flow_count:1 packet_in_count:5 byte_in_count:300 duration:0.0s bands:
+0: packet_count:1 byte_count:60
+])
+
+# Advance time by 1/2 second
+ovs-appctl time/warp 500
+
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p7 'in_port(7),eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800),ipv4(src=10.0.0.2,dst=10.0.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+AT_CHECK([ovs-appctl netdev-dummy/receive p8 'in_port(8),eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800),ipv4(src=10.0.0.3,dst=10.0.0.4,proto=1,tos=0,ttl=64,frag=no),icmp(type=8,code=0)' --len 60])
+sleep 1 # wait for forwarders process packets
+
+# Meter 1 is measuring packets, allowing one packet per second with
+# bursts of one packet, so all 5 of the new packets should hit the drop
+# band.
+# Meter 2 is measuring kbps, with burst size 2 (== 2000 bits). After 500ms
+# there should be space for 80 + 500 bits, so one new 60 byte (480 bit) packet
+# should pass, remaining 4 should hit the drop band.
+AT_CHECK([ovs-ofctl -O OpenFlow13 meter-stats br0 | strip_timers], [0], [dnl
+OFPST_METER reply (OF1.3) (xid=0x2):
+meter:1 flow_count:1 packet_in_count:10 byte_in_count:600 duration:0.0s bands:
+0: packet_count:9 byte_count:540
+
+meter:2 flow_count:1 packet_in_count:10 byte_in_count:600 duration:0.0s bands:
+0: packet_count:5 byte_count:300
+])
+
+AT_CHECK([cat ovs-vswitchd.log | filter_flow_install | strip_xout_keep_actions], [0], [dnl
+recirc_id(0),in_port(1),eth_type(0x0800),ipv4(frag=no), actions:meter(0),7
+recirc_id(0),in_port(2),eth_type(0x0800),ipv4(frag=no), actions:8
+recirc_id(0),in_port(7),eth_type(0x0800),ipv4(frag=no), actions:meter(1),1
+recirc_id(0),in_port(8),eth_type(0x0800),ipv4(frag=no), actions:2
+])
+
+OVS_VSWITCHD_STOP
+AT_CLEANUP
--
1.9.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [userspace meter v3 0/5] Userspace meter
2017-01-31 6:39 [userspace meter v3 0/5] Userspace meter Andy Zhou
` (4 preceding siblings ...)
2017-01-31 6:40 ` [userspace meter v3 5/5] dpif-netdev: Simple DROP meter implementation Andy Zhou
@ 2017-01-31 7:05 ` Andy Zhou
5 siblings, 0 replies; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 7:05 UTC (permalink / raw)
To: David S. Miller; +Cc: Linux Kernel Network Developers, Andy Zhou
Please discard this series. These are ovs user space changes, not
kernel changes. They are sent in error. Sorry.
On Mon, Jan 30, 2017 at 10:39 PM, Andy Zhou <azhou@ovn.org> wrote:
> Repost user space meter support. This is based Jarno's original work
> at: https://mail.openvswitch.org/pipermail/ovs-dev/2015-November/306304.html.
> With some enhancements, and rebased to current master.
>
> ---
> v1-v2: rebase and repost.
> v2-v3: simplify patch 2/5.
>
> Andy Zhou (2):
> netdev-dummy: Add --len option for netdev-dummy/receive command
> dp-packet: Enhance packet batch APIs.
>
> Jarno Rajahalme (3):
> dpif: Meter framework.
> ofproto: Meter translation.
> dpif-netdev: Simple DROP meter implementation.
>
> datapath/linux/compat/include/linux/openvswitch.h | 4 +-
> include/openvswitch/ofp-actions.h | 1 +
> lib/dp-packet.h | 140 +++++--
> lib/dpif-netdev.c | 433 ++++++++++++++++++++--
> lib/dpif-netlink.c | 46 ++-
> lib/dpif-provider.h | 29 ++
> lib/dpif.c | 128 ++++++-
> lib/dpif.h | 13 +-
> lib/netdev-dummy.c | 48 ++-
> lib/netdev-linux.c | 3 +-
> lib/netdev.c | 24 +-
> lib/odp-execute.c | 58 +--
> lib/odp-util.c | 14 +
> lib/ofp-actions.c | 1 +
> ofproto/ofproto-dpif-sflow.c | 1 +
> ofproto/ofproto-dpif-xlate.c | 13 +-
> ofproto/ofproto-dpif.c | 60 ++-
> ofproto/ofproto-provider.h | 13 +-
> ofproto/ofproto.c | 51 +--
> tests/dpif-netdev.at | 106 ++++++
> tests/test-conntrack.c | 56 ++-
> 21 files changed, 1036 insertions(+), 206 deletions(-)
>
> --
> 1.9.1
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [userspace meter v3 1/5] netdev-dummy: Add --len option for netdev-dummy/receive command
2017-01-31 6:39 ` [userspace meter v3 1/5] netdev-dummy: Add --len option for netdev-dummy/receive command Andy Zhou
@ 2017-01-31 7:30 ` Andy Zhou
0 siblings, 0 replies; 8+ messages in thread
From: Andy Zhou @ 2017-01-31 7:30 UTC (permalink / raw)
To: David S. Miller; +Cc: Linux Kernel Network Developers, Andy Zhou
Please discard this series. These are ovs user space changes, not
kernel changes. They are sent in error. Sorry.
On Mon, Jan 30, 2017 at 10:39 PM, Andy Zhou <azhou@ovn.org> wrote:
> Currently, there is no way to specify the packet size when injecting
> a packet via "netdev-dummy/receive" with a flow specification. Thus
> far, packet size is not important for testing OVS features, but it
> becomes useful in writing unit tests for the meter implementation
> in a later patch.
>
> Signed-off-by: Andy Zhou <azhou@ovn.org>
> ---
> lib/netdev-dummy.c | 38 ++++++++++++++++++++++++++++----------
> 1 file changed, 28 insertions(+), 10 deletions(-)
>
> diff --git a/lib/netdev-dummy.c b/lib/netdev-dummy.c
> index e6e36cd..10df0a7 100644
> --- a/lib/netdev-dummy.c
> +++ b/lib/netdev-dummy.c
> @@ -1,5 +1,5 @@
> /*
> - * Copyright (c) 2010, 2011, 2012, 2013, 2015, 2016 Nicira, Inc.
> + * Copyright (c) 2010, 2011, 2012, 2013, 2015, 2016, 2017 Nicira, Inc.
> *
> * Licensed under the Apache License, Version 2.0 (the "License");
> * you may not use this file except in compliance with the License.
> @@ -1433,7 +1433,15 @@ pkt_list_delete(struct ovs_list *l)
> }
>
> static struct dp_packet *
> -eth_from_packet_or_flow(const char *s)
> +eth_from_packet(const char *s)
> +{
> + struct dp_packet *packet;
> + eth_from_hex(s, &packet);
> + return packet;
> +}
> +
> +static struct dp_packet *
> +eth_from_flow(const char *s)
> {
> enum odp_key_fitness fitness;
> struct dp_packet *packet;
> @@ -1441,10 +1449,6 @@ eth_from_packet_or_flow(const char *s)
> struct flow flow;
> int error;
>
> - if (!eth_from_hex(s, &packet)) {
> - return packet;
> - }
> -
> /* Convert string to datapath key.
> *
> * It would actually be nicer to parse an OpenFlow-like flow key here, but
> @@ -1540,10 +1544,24 @@ netdev_dummy_receive(struct unixctl_conn *conn,
> for (i = k; i < argc; i++) {
> struct dp_packet *packet;
>
> - packet = eth_from_packet_or_flow(argv[i]);
> + /* Try to parse 'argv[i]' as packet in hex. */
> + packet = eth_from_packet(argv[i]);
> +
> if (!packet) {
> - unixctl_command_reply_error(conn, "bad packet syntax");
> - goto exit;
> + /* Try parse 'argv[i]' as odp flow. */
> + packet = eth_from_flow(argv[i]);
> +
> + if (!packet) {
> + unixctl_command_reply_error(conn, "bad packet or flow syntax");
> + goto exit;
> + }
> +
> + /* Parse optional --len argument immediately follow a 'flow'. */
> + if (argc >= i + 2 && !strcmp(argv[i + 1], "--len")) {
> + int packet_size = strtol(argv[i + 2], NULL, 10);
> + dp_packet_set_size(packet, packet_size);
> + i+=2;
> + }
> }
>
> netdev_dummy_queue_packet(dummy_dev, packet, rx_qid);
> @@ -1757,7 +1775,7 @@ void
> netdev_dummy_register(enum dummy_level level)
> {
> unixctl_command_register("netdev-dummy/receive",
> - "name [--qid queue_id] packet|flow...",
> + "name [--qid queue_id] packet|flow [--len packet len]..",
> 2, INT_MAX, netdev_dummy_receive, NULL);
> unixctl_command_register("netdev-dummy/set-admin-state",
> "[netdev] up|down", 1, 2,
> --
> 1.9.1
>
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2017-01-31 7:31 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-01-31 6:39 [userspace meter v3 0/5] Userspace meter Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 1/5] netdev-dummy: Add --len option for netdev-dummy/receive command Andy Zhou
2017-01-31 7:30 ` Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 2/5] dp-packet: Enhance packet batch APIs Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 3/5] dpif: Meter framework Andy Zhou
2017-01-31 6:39 ` [userspace meter v3 4/5] ofproto: Meter translation Andy Zhou
2017-01-31 6:40 ` [userspace meter v3 5/5] dpif-netdev: Simple DROP meter implementation Andy Zhou
2017-01-31 7:05 ` [userspace meter v3 0/5] Userspace meter Andy Zhou
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.