2
0
mirror of https://github.com/openvswitch/ovs synced 2025-08-22 18:07:40 +00:00
ovs/lib/dpif.c

1536 lines
49 KiB
C
Raw Normal View History

/*
* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <config.h>
#include "dpif-provider.h"
#include <ctype.h>
#include <errno.h>
#include <inttypes.h>
#include <stdlib.h>
#include <string.h>
#include "coverage.h"
#include "dynamic-string.h"
#include "flow.h"
#include "netdev.h"
#include "netlink.h"
#include "odp-execute.h"
#include "odp-util.h"
#include "ofp-errors.h"
#include "ofp-print.h"
#include "ofp-util.h"
#include "ofpbuf.h"
#include "packets.h"
#include "poll-loop.h"
#include "shash.h"
#include "sset.h"
#include "timeval.h"
#include "util.h"
#include "valgrind.h"
#include "vlog.h"
VLOG_DEFINE_THIS_MODULE(dpif);
COVERAGE_DEFINE(dpif_destroy);
COVERAGE_DEFINE(dpif_port_add);
COVERAGE_DEFINE(dpif_port_del);
COVERAGE_DEFINE(dpif_flow_flush);
COVERAGE_DEFINE(dpif_flow_get);
COVERAGE_DEFINE(dpif_flow_put);
COVERAGE_DEFINE(dpif_flow_del);
COVERAGE_DEFINE(dpif_execute);
COVERAGE_DEFINE(dpif_purge);
COVERAGE_DEFINE(dpif_execute_with_help);
static const struct dpif_class *base_dpif_classes[] = {
#ifdef __linux__
&dpif_linux_class,
#endif
&dpif_netdev_class,
};
struct registered_dpif_class {
const struct dpif_class *dpif_class;
int refcount;
};
static struct shash dpif_classes = SHASH_INITIALIZER(&dpif_classes);
static struct sset dpif_blacklist = SSET_INITIALIZER(&dpif_blacklist);
/* Protects 'dpif_classes', including the refcount, and 'dpif_blacklist'. */
static struct ovs_mutex dpif_mutex = OVS_MUTEX_INITIALIZER;
/* Rate limit for individual messages going to or from the datapath, output at
* DBG level. This is very high because, if these are enabled, it is because
* we really need to see them. */
static struct vlog_rate_limit dpmsg_rl = VLOG_RATE_LIMIT_INIT(600, 600);
/* Not really much point in logging many dpif errors. */
static struct vlog_rate_limit error_rl = VLOG_RATE_LIMIT_INIT(60, 5);
static void log_flow_message(const struct dpif *dpif, int error,
const char *operation,
const struct nlattr *key, size_t key_len,
const struct nlattr *mask, size_t mask_len,
const struct dpif_flow_stats *stats,
const struct nlattr *actions, size_t actions_len);
static void log_operation(const struct dpif *, const char *operation,
int error);
static bool should_log_flow_message(int error);
static void log_flow_put_message(struct dpif *, const struct dpif_flow_put *,
int error);
static void log_flow_del_message(struct dpif *, const struct dpif_flow_del *,
int error);
static void log_execute_message(struct dpif *, const struct dpif_execute *,
int error);
static void
dp_initialize(void)
{
static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
if (ovsthread_once_start(&once)) {
int i;
for (i = 0; i < ARRAY_SIZE(base_dpif_classes); i++) {
dp_register_provider(base_dpif_classes[i]);
}
ovsthread_once_done(&once);
}
}
static int
dp_register_provider__(const struct dpif_class *new_class)
{
struct registered_dpif_class *registered_class;
if (sset_contains(&dpif_blacklist, new_class->type)) {
VLOG_DBG("attempted to register blacklisted provider: %s",
new_class->type);
return EINVAL;
}
if (shash_find(&dpif_classes, new_class->type)) {
VLOG_WARN("attempted to register duplicate datapath provider: %s",
new_class->type);
return EEXIST;
}
registered_class = xmalloc(sizeof *registered_class);
registered_class->dpif_class = new_class;
registered_class->refcount = 0;
shash_add(&dpif_classes, new_class->type, registered_class);
return 0;
}
/* Registers a new datapath provider. After successful registration, new
* datapaths of that type can be opened using dpif_open(). */
int
dp_register_provider(const struct dpif_class *new_class)
{
int error;
ovs_mutex_lock(&dpif_mutex);
error = dp_register_provider__(new_class);
ovs_mutex_unlock(&dpif_mutex);
return error;
}
/* Unregisters a datapath provider. 'type' must have been previously
* registered and not currently be in use by any dpifs. After unregistration
* new datapaths of that type cannot be opened using dpif_open(). */
static int
dp_unregister_provider__(const char *type)
{
struct shash_node *node;
struct registered_dpif_class *registered_class;
node = shash_find(&dpif_classes, type);
if (!node) {
VLOG_WARN("attempted to unregister a datapath provider that is not "
"registered: %s", type);
return EAFNOSUPPORT;
}
registered_class = node->data;
if (registered_class->refcount) {
VLOG_WARN("attempted to unregister in use datapath provider: %s", type);
return EBUSY;
}
shash_delete(&dpif_classes, node);
free(registered_class);
return 0;
}
/* Unregisters a datapath provider. 'type' must have been previously
* registered and not currently be in use by any dpifs. After unregistration
* new datapaths of that type cannot be opened using dpif_open(). */
int
dp_unregister_provider(const char *type)
{
int error;
dp_initialize();
ovs_mutex_lock(&dpif_mutex);
error = dp_unregister_provider__(type);
ovs_mutex_unlock(&dpif_mutex);
return error;
}
/* Blacklists a provider. Causes future calls of dp_register_provider() with
* a dpif_class which implements 'type' to fail. */
void
dp_blacklist_provider(const char *type)
{
ovs_mutex_lock(&dpif_mutex);
sset_add(&dpif_blacklist, type);
ovs_mutex_unlock(&dpif_mutex);
}
/* Clears 'types' and enumerates the types of all currently registered datapath
* providers into it. The caller must first initialize the sset. */
void
dp_enumerate_types(struct sset *types)
{
struct shash_node *node;
dp_initialize();
sset_clear(types);
ovs_mutex_lock(&dpif_mutex);
SHASH_FOR_EACH(node, &dpif_classes) {
const struct registered_dpif_class *registered_class = node->data;
sset_add(types, registered_class->dpif_class->type);
}
ovs_mutex_unlock(&dpif_mutex);
}
static void
dp_class_unref(struct registered_dpif_class *rc)
{
ovs_mutex_lock(&dpif_mutex);
ovs_assert(rc->refcount);
rc->refcount--;
ovs_mutex_unlock(&dpif_mutex);
}
static struct registered_dpif_class *
dp_class_lookup(const char *type)
{
struct registered_dpif_class *rc;
ovs_mutex_lock(&dpif_mutex);
rc = shash_find_data(&dpif_classes, type);
if (rc) {
rc->refcount++;
}
ovs_mutex_unlock(&dpif_mutex);
return rc;
}
/* Clears 'names' and enumerates the names of all known created datapaths with
* the given 'type'. The caller must first initialize the sset. Returns 0 if
* successful, otherwise a positive errno value.
*
* Some kinds of datapaths might not be practically enumerable. This is not
* considered an error. */
int
dp_enumerate_names(const char *type, struct sset *names)
{
struct registered_dpif_class *registered_class;
const struct dpif_class *dpif_class;
int error;
dp_initialize();
sset_clear(names);
registered_class = dp_class_lookup(type);
if (!registered_class) {
VLOG_WARN("could not enumerate unknown type: %s", type);
return EAFNOSUPPORT;
}
dpif_class = registered_class->dpif_class;
error = dpif_class->enumerate ? dpif_class->enumerate(names) : 0;
if (error) {
VLOG_WARN("failed to enumerate %s datapaths: %s", dpif_class->type,
ovs_strerror(error));
}
dp_class_unref(registered_class);
return error;
}
/* Parses 'datapath_name_', which is of the form [type@]name into its
* component pieces. 'name' and 'type' must be freed by the caller.
*
* The returned 'type' is normalized, as if by dpif_normalize_type(). */
void
dp_parse_name(const char *datapath_name_, char **name, char **type)
{
char *datapath_name = xstrdup(datapath_name_);
char *separator;
separator = strchr(datapath_name, '@');
if (separator) {
*separator = '\0';
*type = datapath_name;
*name = xstrdup(dpif_normalize_type(separator + 1));
} else {
*name = datapath_name;
*type = xstrdup(dpif_normalize_type(NULL));
}
}
static int
do_open(const char *name, const char *type, bool create, struct dpif **dpifp)
{
struct dpif *dpif = NULL;
int error;
struct registered_dpif_class *registered_class;
dp_initialize();
type = dpif_normalize_type(type);
registered_class = dp_class_lookup(type);
if (!registered_class) {
VLOG_WARN("could not create datapath %s of unknown type %s", name,
type);
error = EAFNOSUPPORT;
goto exit;
}
error = registered_class->dpif_class->open(registered_class->dpif_class,
name, create, &dpif);
if (!error) {
ovs_assert(dpif->dpif_class == registered_class->dpif_class);
} else {
dp_class_unref(registered_class);
}
exit:
*dpifp = error ? NULL : dpif;
return error;
}
/* Tries to open an existing datapath named 'name' and type 'type'. Will fail
* if no datapath with 'name' and 'type' exists. 'type' may be either NULL or
* the empty string to specify the default system type. Returns 0 if
* successful, otherwise a positive errno value. On success stores a pointer
* to the datapath in '*dpifp', otherwise a null pointer. */
int
dpif_open(const char *name, const char *type, struct dpif **dpifp)
{
return do_open(name, type, false, dpifp);
}
/* Tries to create and open a new datapath with the given 'name' and 'type'.
* 'type' may be either NULL or the empty string to specify the default system
* type. Will fail if a datapath with 'name' and 'type' already exists.
* Returns 0 if successful, otherwise a positive errno value. On success
* stores a pointer to the datapath in '*dpifp', otherwise a null pointer. */
int
dpif_create(const char *name, const char *type, struct dpif **dpifp)
{
return do_open(name, type, true, dpifp);
}
/* Tries to open a datapath with the given 'name' and 'type', creating it if it
* does not exist. 'type' may be either NULL or the empty string to specify
* the default system type. Returns 0 if successful, otherwise a positive
* errno value. On success stores a pointer to the datapath in '*dpifp',
* otherwise a null pointer. */
int
dpif_create_and_open(const char *name, const char *type, struct dpif **dpifp)
{
int error;
error = dpif_create(name, type, dpifp);
if (error == EEXIST || error == EBUSY) {
error = dpif_open(name, type, dpifp);
if (error) {
VLOG_WARN("datapath %s already exists but cannot be opened: %s",
name, ovs_strerror(error));
}
} else if (error) {
VLOG_WARN("failed to create datapath %s: %s",
name, ovs_strerror(error));
}
return error;
}
/* Closes and frees the connection to 'dpif'. Does not destroy the datapath
* itself; call dpif_delete() first, instead, if that is desirable. */
void
dpif_close(struct dpif *dpif)
{
if (dpif) {
struct registered_dpif_class *rc;
rc = shash_find_data(&dpif_classes, dpif->dpif_class->type);
dpif_uninit(dpif, true);
dp_class_unref(rc);
}
}
/* Performs periodic work needed by 'dpif'. */
void
dpif_run(struct dpif *dpif)
{
if (dpif->dpif_class->run) {
dpif->dpif_class->run(dpif);
}
}
/* Arranges for poll_block() to wake up when dp_run() needs to be called for
* 'dpif'. */
void
dpif_wait(struct dpif *dpif)
{
if (dpif->dpif_class->wait) {
dpif->dpif_class->wait(dpif);
}
}
/* Returns the name of datapath 'dpif' prefixed with the type
* (for use in log messages). */
const char *
dpif_name(const struct dpif *dpif)
{
return dpif->full_name;
}
/* Returns the name of datapath 'dpif' without the type
* (for use in device names). */
const char *
dpif_base_name(const struct dpif *dpif)
{
return dpif->base_name;
}
/* Returns the type of datapath 'dpif'. */
const char *
dpif_type(const struct dpif *dpif)
{
return dpif->dpif_class->type;
}
/* Returns the fully spelled out name for the given datapath 'type'.
*
* Normalized type string can be compared with strcmp(). Unnormalized type
* string might be the same even if they have different spellings. */
const char *
dpif_normalize_type(const char *type)
{
return type && type[0] ? type : "system";
}
/* Destroys the datapath that 'dpif' is connected to, first removing all of its
* ports. After calling this function, it does not make sense to pass 'dpif'
* to any functions other than dpif_name() or dpif_close(). */
int
dpif_delete(struct dpif *dpif)
{
int error;
COVERAGE_INC(dpif_destroy);
error = dpif->dpif_class->destroy(dpif);
log_operation(dpif, "delete", error);
return error;
}
/* Retrieves statistics for 'dpif' into 'stats'. Returns 0 if successful,
* otherwise a positive errno value. */
int
dpif_get_dp_stats(const struct dpif *dpif, struct dpif_dp_stats *stats)
{
int error = dpif->dpif_class->get_stats(dpif, stats);
if (error) {
memset(stats, 0, sizeof *stats);
}
log_operation(dpif, "get_stats", error);
return error;
}
const char *
dpif_port_open_type(const char *datapath_type, const char *port_type)
{
struct registered_dpif_class *rc;
datapath_type = dpif_normalize_type(datapath_type);
ovs_mutex_lock(&dpif_mutex);
rc = shash_find_data(&dpif_classes, datapath_type);
if (rc && rc->dpif_class->port_open_type) {
port_type = rc->dpif_class->port_open_type(rc->dpif_class, port_type);
}
ovs_mutex_unlock(&dpif_mutex);
return port_type;
}
/* Attempts to add 'netdev' as a port on 'dpif'. If 'port_nop' is
* non-null and its value is not ODPP_NONE, then attempts to use the
* value as the port number.
*
* If successful, returns 0 and sets '*port_nop' to the new port's port
* number (if 'port_nop' is non-null). On failure, returns a positive
* errno value and sets '*port_nop' to ODPP_NONE (if 'port_nop' is
* non-null). */
int
dpif_port_add(struct dpif *dpif, struct netdev *netdev, odp_port_t *port_nop)
{
const char *netdev_name = netdev_get_name(netdev);
odp_port_t port_no = ODPP_NONE;
int error;
COVERAGE_INC(dpif_port_add);
if (port_nop) {
port_no = *port_nop;
}
error = dpif->dpif_class->port_add(dpif, netdev, &port_no);
if (!error) {
VLOG_DBG_RL(&dpmsg_rl, "%s: added %s as port %"PRIu32,
dpif_name(dpif), netdev_name, port_no);
} else {
VLOG_WARN_RL(&error_rl, "%s: failed to add %s as port: %s",
dpif_name(dpif), netdev_name, ovs_strerror(error));
port_no = ODPP_NONE;
}
if (port_nop) {
*port_nop = port_no;
}
return error;
}
/* Attempts to remove 'dpif''s port number 'port_no'. Returns 0 if successful,
* otherwise a positive errno value. */
int
dpif_port_del(struct dpif *dpif, odp_port_t port_no)
{
int error;
COVERAGE_INC(dpif_port_del);
error = dpif->dpif_class->port_del(dpif, port_no);
if (!error) {
VLOG_DBG_RL(&dpmsg_rl, "%s: port_del(%"PRIu32")",
dpif_name(dpif), port_no);
} else {
log_operation(dpif, "port_del", error);
}
return error;
}
/* Makes a deep copy of 'src' into 'dst'. */
void
dpif_port_clone(struct dpif_port *dst, const struct dpif_port *src)
{
dst->name = xstrdup(src->name);
dst->type = xstrdup(src->type);
dst->port_no = src->port_no;
}
/* Frees memory allocated to members of 'dpif_port'.
*
* Do not call this function on a dpif_port obtained from
* dpif_port_dump_next(): that function retains ownership of the data in the
* dpif_port. */
void
dpif_port_destroy(struct dpif_port *dpif_port)
{
free(dpif_port->name);
free(dpif_port->type);
}
/* Checks if port named 'devname' exists in 'dpif'. If so, returns
* true; otherwise, returns false. */
bool
dpif_port_exists(const struct dpif *dpif, const char *devname)
{
int error = dpif->dpif_class->port_query_by_name(dpif, devname, NULL);
if (error != 0 && error != ENOENT && error != ENODEV) {
VLOG_WARN_RL(&error_rl, "%s: failed to query port %s: %s",
dpif_name(dpif), devname, ovs_strerror(error));
}
return !error;
}
/* Looks up port number 'port_no' in 'dpif'. On success, returns 0 and
* initializes '*port' appropriately; on failure, returns a positive errno
* value.
*
* The caller owns the data in 'port' and must free it with
* dpif_port_destroy() when it is no longer needed. */
int
dpif_port_query_by_number(const struct dpif *dpif, odp_port_t port_no,
struct dpif_port *port)
{
int error = dpif->dpif_class->port_query_by_number(dpif, port_no, port);
if (!error) {
VLOG_DBG_RL(&dpmsg_rl, "%s: port %"PRIu32" is device %s",
dpif_name(dpif), port_no, port->name);
} else {
memset(port, 0, sizeof *port);
VLOG_WARN_RL(&error_rl, "%s: failed to query port %"PRIu32": %s",
dpif_name(dpif), port_no, ovs_strerror(error));
}
return error;
}
/* Looks up port named 'devname' in 'dpif'. On success, returns 0 and
* initializes '*port' appropriately; on failure, returns a positive errno
* value.
*
* The caller owns the data in 'port' and must free it with
* dpif_port_destroy() when it is no longer needed. */
int
dpif_port_query_by_name(const struct dpif *dpif, const char *devname,
struct dpif_port *port)
{
int error = dpif->dpif_class->port_query_by_name(dpif, devname, port);
if (!error) {
VLOG_DBG_RL(&dpmsg_rl, "%s: device %s is on port %"PRIu32,
dpif_name(dpif), devname, port->port_no);
} else {
memset(port, 0, sizeof *port);
/* For ENOENT or ENODEV we use DBG level because the caller is probably
* interested in whether 'dpif' actually has a port 'devname', so that
* it's not an issue worth logging if it doesn't. Other errors are
* uncommon and more likely to indicate a real problem. */
VLOG_RL(&error_rl,
error == ENOENT || error == ENODEV ? VLL_DBG : VLL_WARN,
"%s: failed to query port %s: %s",
dpif_name(dpif), devname, ovs_strerror(error));
}
return error;
}
/* Returns the Netlink PID value to supply in OVS_ACTION_ATTR_USERSPACE actions
* as the OVS_USERSPACE_ATTR_PID attribute's value, for use in flows whose
* packets arrived on port 'port_no'.
*
* A 'port_no' of ODPP_NONE is a special case: it returns a reserved PID, not
* allocated to any port, that the client may use for special purposes.
*
* The return value is only meaningful when DPIF_UC_ACTION has been enabled in
* the 'dpif''s listen mask. It is allowed to change when DPIF_UC_ACTION is
* disabled and then re-enabled, so a client that does that must be prepared to
* update all of the flows that it installed that contain
* OVS_ACTION_ATTR_USERSPACE actions. */
uint32_t
dpif_port_get_pid(const struct dpif *dpif, odp_port_t port_no)
{
return (dpif->dpif_class->port_get_pid
? (dpif->dpif_class->port_get_pid)(dpif, port_no)
: 0);
}
/* Looks up port number 'port_no' in 'dpif'. On success, returns 0 and copies
* the port's name into the 'name_size' bytes in 'name', ensuring that the
* result is null-terminated. On failure, returns a positive errno value and
* makes 'name' the empty string. */
int
dpif_port_get_name(struct dpif *dpif, odp_port_t port_no,
char *name, size_t name_size)
{
struct dpif_port port;
int error;
ovs_assert(name_size > 0);
error = dpif_port_query_by_number(dpif, port_no, &port);
if (!error) {
ovs_strlcpy(name, port.name, name_size);
dpif_port_destroy(&port);
} else {
*name = '\0';
}
return error;
}
2011-01-10 13:12:12 -08:00
/* Initializes 'dump' to begin dumping the ports in a dpif.
*
2011-01-10 13:12:12 -08:00
* This function provides no status indication. An error status for the entire
* dump operation is provided when it is completed by calling
* dpif_port_dump_done().
*/
void
dpif_port_dump_start(struct dpif_port_dump *dump, const struct dpif *dpif)
{
dump->dpif = dpif;
dump->error = dpif->dpif_class->port_dump_start(dpif, &dump->state);
log_operation(dpif, "port_dump_start", dump->error);
}
/* Attempts to retrieve another port from 'dump', which must have been
* initialized with dpif_port_dump_start(). On success, stores a new dpif_port
2011-01-10 13:12:12 -08:00
* into 'port' and returns true. On failure, returns false.
*
2011-01-10 13:12:12 -08:00
* Failure might indicate an actual error or merely that the last port has been
* dumped. An error status for the entire dump operation is provided when it
* is completed by calling dpif_port_dump_done().
*
* The dpif owns the data stored in 'port'. It will remain valid until at
* least the next time 'dump' is passed to dpif_port_dump_next() or
* dpif_port_dump_done(). */
2011-01-10 13:12:12 -08:00
bool
dpif_port_dump_next(struct dpif_port_dump *dump, struct dpif_port *port)
{
2011-01-10 13:12:12 -08:00
const struct dpif *dpif = dump->dpif;
2011-01-10 13:12:12 -08:00
if (dump->error) {
return false;
}
2011-01-10 13:12:12 -08:00
dump->error = dpif->dpif_class->port_dump_next(dpif, dump->state, port);
if (dump->error == EOF) {
VLOG_DBG_RL(&dpmsg_rl, "%s: dumped all ports", dpif_name(dpif));
} else {
log_operation(dpif, "port_dump_next", dump->error);
}
2011-01-10 13:12:12 -08:00
if (dump->error) {
dpif->dpif_class->port_dump_done(dpif, dump->state);
return false;
}
2011-01-10 13:12:12 -08:00
return true;
}
2011-01-10 13:12:12 -08:00
/* Completes port table dump operation 'dump', which must have been initialized
* with dpif_port_dump_start(). Returns 0 if the dump operation was
* error-free, otherwise a positive errno value describing the problem. */
int
dpif_port_dump_done(struct dpif_port_dump *dump)
{
const struct dpif *dpif = dump->dpif;
if (!dump->error) {
dump->error = dpif->dpif_class->port_dump_done(dpif, dump->state);
log_operation(dpif, "port_dump_done", dump->error);
}
2011-01-10 13:12:12 -08:00
return dump->error == EOF ? 0 : dump->error;
}
/* Polls for changes in the set of ports in 'dpif'. If the set of ports in
* 'dpif' has changed, this function does one of the following:
*
* - Stores the name of the device that was added to or deleted from 'dpif' in
* '*devnamep' and returns 0. The caller is responsible for freeing
* '*devnamep' (with free()) when it no longer needs it.
*
* - Returns ENOBUFS and sets '*devnamep' to NULL.
*
* This function may also return 'false positives', where it returns 0 and
* '*devnamep' names a device that was not actually added or deleted or it
* returns ENOBUFS without any change.
*
* Returns EAGAIN if the set of ports in 'dpif' has not changed. May also
* return other positive errno values to indicate that something has gone
* wrong. */
int
dpif_port_poll(const struct dpif *dpif, char **devnamep)
{
int error = dpif->dpif_class->port_poll(dpif, devnamep);
if (error) {
*devnamep = NULL;
}
return error;
}
/* Arranges for the poll loop to wake up when port_poll(dpif) will return a
* value other than EAGAIN. */
void
dpif_port_poll_wait(const struct dpif *dpif)
{
dpif->dpif_class->port_poll_wait(dpif);
}
/* Extracts the flow stats for a packet. The 'flow' and 'packet'
* arguments must have been initialized through a call to flow_extract().
* 'used' is stored into stats->used. */
void
dpif_flow_stats_extract(const struct flow *flow, const struct ofpbuf *packet,
long long int used, struct dpif_flow_stats *stats)
{
stats->tcp_flags = packet_get_tcp_flags(packet, flow);
stats->n_bytes = packet->size;
stats->n_packets = 1;
stats->used = used;
}
/* Appends a human-readable representation of 'stats' to 's'. */
void
dpif_flow_stats_format(const struct dpif_flow_stats *stats, struct ds *s)
{
ds_put_format(s, "packets:%"PRIu64", bytes:%"PRIu64", used:",
stats->n_packets, stats->n_bytes);
if (stats->used) {
ds_put_format(s, "%.3fs", (time_msec() - stats->used) / 1000.0);
} else {
ds_put_format(s, "never");
}
if (stats->tcp_flags) {
ds_put_cstr(s, ", flags:");
packet_format_tcp_flags(s, stats->tcp_flags);
}
}
/* Deletes all flows from 'dpif'. Returns 0 if successful, otherwise a
* positive errno value. */
int
dpif_flow_flush(struct dpif *dpif)
{
int error;
COVERAGE_INC(dpif_flow_flush);
error = dpif->dpif_class->flow_flush(dpif);
log_operation(dpif, "flow_flush", error);
return error;
}
/* Queries 'dpif' for a flow entry. The flow is specified by the Netlink
* attributes with types OVS_KEY_ATTR_* in the 'key_len' bytes starting at
* 'key'.
*
* Returns 0 if successful. If no flow matches, returns ENOENT. On other
* failure, returns a positive errno value.
*
* If 'actionsp' is nonnull, then on success '*actionsp' will be set to an
* ofpbuf owned by the caller that contains the Netlink attributes for the
* flow's actions. The caller must free the ofpbuf (with ofpbuf_delete()) when
* it is no longer needed.
*
* If 'stats' is nonnull, then on success it will be updated with the flow's
* statistics. */
int
dpif_flow_get(const struct dpif *dpif,
const struct nlattr *key, size_t key_len,
struct ofpbuf **actionsp, struct dpif_flow_stats *stats)
{
int error;
COVERAGE_INC(dpif_flow_get);
error = dpif->dpif_class->flow_get(dpif, key, key_len, actionsp, stats);
if (error) {
if (actionsp) {
*actionsp = NULL;
}
if (stats) {
memset(stats, 0, sizeof *stats);
}
}
if (should_log_flow_message(error)) {
const struct nlattr *actions;
size_t actions_len;
if (!error && actionsp) {
actions = (*actionsp)->data;
actions_len = (*actionsp)->size;
} else {
actions = NULL;
actions_len = 0;
}
log_flow_message(dpif, error, "flow_get", key, key_len,
NULL, 0, stats, actions, actions_len);
}
return error;
}
static int
dpif_flow_put__(struct dpif *dpif, const struct dpif_flow_put *put)
{
int error;
COVERAGE_INC(dpif_flow_put);
ovs_assert(!(put->flags & ~(DPIF_FP_CREATE | DPIF_FP_MODIFY
| DPIF_FP_ZERO_STATS)));
error = dpif->dpif_class->flow_put(dpif, put);
if (error && put->stats) {
memset(put->stats, 0, sizeof *put->stats);
}
log_flow_put_message(dpif, put, error);
return error;
}
/* Adds or modifies a flow in 'dpif'. The flow is specified by the Netlink
* attribute OVS_FLOW_ATTR_KEY with types OVS_KEY_ATTR_* in the 'key_len' bytes
* starting at 'key', and OVS_FLOW_ATTR_MASK with types of OVS_KEY_ATTR_* in
* the 'mask_len' bytes starting at 'mask'. The associated actions are
* specified by the Netlink attributes with types OVS_ACTION_ATTR_* in the
* 'actions_len' bytes starting at 'actions'.
*
* - If the flow's key does not exist in 'dpif', then the flow will be added if
* 'flags' includes DPIF_FP_CREATE. Otherwise the operation will fail with
* ENOENT.
*
* The datapath may reject attempts to insert overlapping flows with EINVAL
* or EEXIST, but clients should not rely on this: avoiding overlapping flows
* is primarily the client's responsibility.
*
* If the operation succeeds, then 'stats', if nonnull, will be zeroed.
*
* - If the flow's key does exist in 'dpif', then the flow's actions will be
* updated if 'flags' includes DPIF_FP_MODIFY. Otherwise the operation will
* fail with EEXIST. If the flow's actions are updated, then its statistics
* will be zeroed if 'flags' includes DPIF_FP_ZERO_STATS, and left as-is
* otherwise.
*
* If the operation succeeds, then 'stats', if nonnull, will be set to the
* flow's statistics before the update.
*/
int
dpif_flow_put(struct dpif *dpif, enum dpif_flow_put_flags flags,
const struct nlattr *key, size_t key_len,
const struct nlattr *mask, size_t mask_len,
const struct nlattr *actions, size_t actions_len,
struct dpif_flow_stats *stats)
{
struct dpif_flow_put put;
put.flags = flags;
put.key = key;
put.key_len = key_len;
put.mask = mask;
put.mask_len = mask_len;
put.actions = actions;
put.actions_len = actions_len;
put.stats = stats;
return dpif_flow_put__(dpif, &put);
}
static int
dpif_flow_del__(struct dpif *dpif, struct dpif_flow_del *del)
{
int error;
COVERAGE_INC(dpif_flow_del);
error = dpif->dpif_class->flow_del(dpif, del);
if (error && del->stats) {
memset(del->stats, 0, sizeof *del->stats);
}
log_flow_del_message(dpif, del, error);
return error;
}
/* Deletes a flow from 'dpif' and returns 0, or returns ENOENT if 'dpif' does
* not contain such a flow. The flow is specified by the Netlink attributes
* with types OVS_KEY_ATTR_* in the 'key_len' bytes starting at 'key'.
*
* If the operation succeeds, then 'stats', if nonnull, will be set to the
* flow's statistics before its deletion. */
int
dpif_flow_del(struct dpif *dpif,
const struct nlattr *key, size_t key_len,
struct dpif_flow_stats *stats)
{
struct dpif_flow_del del;
del.key = key;
del.key_len = key_len;
del.stats = stats;
return dpif_flow_del__(dpif, &del);
}
/* Allocates thread-local state for use with the 'flow_dump_next' function for
* 'dpif'. On return, initializes '*statep' with any private data needed for
* iteration. */
void
dpif_flow_dump_state_init(const struct dpif *dpif, void **statep)
{
dpif->dpif_class->flow_dump_state_init(statep);
}
/* Releases 'state' which was initialized by a call to the
* 'flow_dump_state_init' function for 'dpif'. */
void
dpif_flow_dump_state_uninit(const struct dpif *dpif, void *state)
{
dpif->dpif_class->flow_dump_state_uninit(state);
}
/* Initializes 'dump' to begin dumping the flows in a dpif. On sucess,
* initializes 'dump' with any data needed for iteration and returns 0.
* Otherwise, returns a positive errno value describing the problem. */
int
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
dpif_flow_dump_start(struct dpif_flow_dump *dump, const struct dpif *dpif)
{
int error;
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
dump->dpif = dpif;
error = dpif->dpif_class->flow_dump_start(dpif, &dump->iter);
log_operation(dpif, "flow_dump_start", error);
return error;
}
/* Attempts to retrieve another flow from 'dump', using 'state' for
* thread-local storage. 'dump' must have been initialized with a successful
* call to dpif_flow_dump_start(), and 'state' must have been initialized with
* dpif_flow_state_init().
*
* On success, updates the output parameters as described below and returns
* true. Otherwise, returns false. Failure might indicate an actual error or
* merely the end of the flow table. An error status for the entire dump
* operation is provided when it is completed by calling dpif_flow_dump_done().
* Multiple threads may use the same 'dump' with this function, but all other
* parameters must not be shared.
*
* On success, if 'key' and 'key_len' are nonnull then '*key' and '*key_len'
* will be set to Netlink attributes with types OVS_KEY_ATTR_* representing the
* dumped flow's key. If 'actions' and 'actions_len' are nonnull then they are
* set to Netlink attributes with types OVS_ACTION_ATTR_* representing the
* dumped flow's actions. If 'stats' is nonnull then it will be set to the
* dumped flow's statistics.
*
* All of the returned data is owned by 'dpif', not by the caller, and the
* caller must not modify or free it. 'dpif' guarantees that it remains
* accessible and unchanging until at least the next call to 'flow_dump_next'
* or 'flow_dump_done' for 'dump' and 'state'. */
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
bool
dpif_flow_dump_next(struct dpif_flow_dump *dump, void *state,
const struct nlattr **key, size_t *key_len,
const struct nlattr **mask, size_t *mask_len,
const struct nlattr **actions, size_t *actions_len,
const struct dpif_flow_stats **stats)
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
{
const struct dpif *dpif = dump->dpif;
int error;
error = dpif->dpif_class->flow_dump_next(dpif, dump->iter, state,
key, key_len, mask, mask_len,
actions, actions_len, stats);
if (error) {
if (key) {
*key = NULL;
*key_len = 0;
}
if (mask) {
*mask = NULL;
*mask_len = 0;
}
if (actions) {
*actions = NULL;
*actions_len = 0;
}
if (stats) {
*stats = NULL;
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
}
}
if (error == EOF) {
VLOG_DBG_RL(&dpmsg_rl, "%s: dumped all flows", dpif_name(dpif));
} else if (should_log_flow_message(error)) {
log_flow_message(dpif, error, "flow_dump",
key ? *key : NULL, key ? *key_len : 0,
mask ? *mask : NULL, mask ? *mask_len : 0,
stats ? *stats : NULL, actions ? *actions : NULL,
actions ? *actions_len : 0);
}
return !error;
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
}
/* Determines whether the next call to 'dpif_flow_dump_next' for 'dump' and
* 'state' will modify or free the keys that it previously returned. 'state'
* must have been initialized by a call to 'dpif_flow_dump_state_init' for
* 'dump'.
*
* 'dpif' guarantees that data returned by flow_dump_next() will remain
* accessible and unchanging until the next call. This function provides a way
* for callers to determine whether that guarantee extends beyond the next
* call.
*
* Returns true if the next call to flow_dump_next() is expected to be
* destructive to previously returned keys for 'state', false otherwise. */
bool
dpif_flow_dump_next_may_destroy_keys(struct dpif_flow_dump *dump, void *state)
{
const struct dpif *dpif = dump->dpif;
return (dpif->dpif_class->flow_dump_next_may_destroy_keys
? dpif->dpif_class->flow_dump_next_may_destroy_keys(state)
: true);
}
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
/* Completes flow table dump operation 'dump', which must have been initialized
* with a successful call to dpif_flow_dump_start(). Returns 0 if the dump
* operation was error-free, otherwise a positive errno value describing the
* problem. */
datapath: Change listing flows to use an iterator concept. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. In turn, that means that flow keys must become variable-length. This does not, however, fit in well with the ODP_FLOW_LIST ioctl in its current form, because that would require userspace to know how much space to allocate for each flow's key in advance, or to allocate as much space as could possibly be needed. Neither choice is very attractive. This commit prepares for a different solution, by replacing ODP_FLOW_LIST by a new ioctl ODP_FLOW_DUMP that retrieves a single flow from the datapath on each call. It is much cleaner to allocate the maximum amount of space for a single flow key than to do so for possibly a very large number of flow keys. As a side effect, this patch also fixes a race condition that sometimes made "ovs-dpctl dump-flows" print an error: previously, flows were listed and then their actions were retrieved, which left a window in which ovs-vswitchd could delete the flow. Now dumping a flow and its actions is a single step, closing that window. Dumping all of the flows in a datapath is no longer an atomic step, so now it is possible to miss some flows or see a single flow twice during iteration, if the flow table is modified by another process. It doesn't look like this should be a problem for ovs-vswitchd. It would be faster to retrieve a number of flows in batch instead of just one at a time, but that will naturally happen later when the kernel datapath interface is changed to use Netlink, so this patch does not bother with it. Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2010-12-28 10:39:52 -08:00
int
dpif_flow_dump_done(struct dpif_flow_dump *dump)
{
const struct dpif *dpif = dump->dpif;
int error = dpif->dpif_class->flow_dump_done(dpif, dump->iter);
log_operation(dpif, "flow_dump_done", error);
return error == EOF ? 0 : error;
}
struct dpif_execute_helper_aux {
struct dpif *dpif;
int error;
};
/* This is called for actions that need the context of the datapath to be
* meaningful. */
static void
dpif_execute_helper_cb(void *aux_, struct ofpbuf *packet,
const struct pkt_metadata *md,
const struct nlattr *action, bool may_steal OVS_UNUSED)
{
struct dpif_execute_helper_aux *aux = aux_;
struct dpif_execute execute;
int type = nl_attr_type(action);
switch ((enum ovs_action_attr)type) {
case OVS_ACTION_ATTR_OUTPUT:
case OVS_ACTION_ATTR_USERSPACE:
execute.actions = action;
execute.actions_len = NLA_ALIGN(action->nla_len);
execute.packet = packet;
execute.md = *md;
execute.needs_help = false;
aux->error = aux->dpif->dpif_class->execute(aux->dpif, &execute);
break;
case OVS_ACTION_ATTR_PUSH_VLAN:
case OVS_ACTION_ATTR_POP_VLAN:
case OVS_ACTION_ATTR_PUSH_MPLS:
case OVS_ACTION_ATTR_POP_MPLS:
case OVS_ACTION_ATTR_SET:
case OVS_ACTION_ATTR_SAMPLE:
case OVS_ACTION_ATTR_UNSPEC:
case __OVS_ACTION_ATTR_MAX:
OVS_NOT_REACHED();
}
}
/* Executes 'execute' by performing most of the actions in userspace and
* passing the fully constructed packets to 'dpif' for output and userspace
* actions.
*
* This helps with actions that a given 'dpif' doesn't implement directly. */
static int
dpif_execute_with_help(struct dpif *dpif, struct dpif_execute *execute)
{
struct dpif_execute_helper_aux aux = {dpif, 0};
COVERAGE_INC(dpif_execute_with_help);
odp_execute_actions(&aux, execute->packet, &execute->md,
execute->actions, execute->actions_len,
dpif_execute_helper_cb);
return aux.error;
}
/* Causes 'dpif' to perform the 'execute->actions_len' bytes of actions in
* 'execute->actions' on the Ethernet frame in 'execute->packet' and on packet
* metadata in 'execute->md'. The implementation is allowed to modify both the
* '*execute->packet' and 'execute->md'.
*
* Some dpif providers do not implement every action. The Linux kernel
* datapath, in particular, does not implement ARP field modification. If
* 'needs_help' is true, the dpif layer executes in userspace all of the
* actions that it can, and for OVS_ACTION_ATTR_OUTPUT and
* OVS_ACTION_ATTR_USERSPACE actions it passes the packet through to the dpif
* implementation.
*
* This works even if 'execute->actions_len' is too long for a Netlink
* attribute.
*
* Returns 0 if successful, otherwise a positive errno value. */
int
dpif_execute(struct dpif *dpif, struct dpif_execute *execute)
{
int error;
COVERAGE_INC(dpif_execute);
if (execute->actions_len > 0) {
error = (execute->needs_help || nl_attr_oversized(execute->actions_len)
? dpif_execute_with_help(dpif, execute)
: dpif->dpif_class->execute(dpif, execute));
} else {
error = 0;
}
log_execute_message(dpif, execute, error);
return error;
}
/* Executes each of the 'n_ops' operations in 'ops' on 'dpif', in the order in
* which they are specified, placing each operation's results in the "output"
* members documented in comments.
*
* This function exists because some datapaths can perform batched operations
* faster than individual operations. */
void
dpif_operate(struct dpif *dpif, struct dpif_op **ops, size_t n_ops)
{
if (dpif->dpif_class->operate) {
while (n_ops > 0) {
size_t chunk;
/* Count 'chunk', the number of ops that can be executed without
* needing any help. Ops that need help should be rare, so we
* expect this to ordinarily be 'n_ops', that is, all the ops. */
for (chunk = 0; chunk < n_ops; chunk++) {
struct dpif_op *op = ops[chunk];
if (op->type == DPIF_OP_EXECUTE && op->u.execute.needs_help) {
break;
}
}
if (chunk) {
/* Execute a chunk full of ops that the dpif provider can
* handle itself, without help. */
size_t i;
dpif->dpif_class->operate(dpif, ops, chunk);
for (i = 0; i < chunk; i++) {
struct dpif_op *op = ops[i];
switch (op->type) {
case DPIF_OP_FLOW_PUT:
log_flow_put_message(dpif, &op->u.flow_put, op->error);
break;
case DPIF_OP_FLOW_DEL:
log_flow_del_message(dpif, &op->u.flow_del, op->error);
break;
case DPIF_OP_EXECUTE:
log_execute_message(dpif, &op->u.execute, op->error);
break;
}
}
ops += chunk;
n_ops -= chunk;
} else {
/* Help the dpif provider to execute one op. */
struct dpif_op *op = ops[0];
op->error = dpif_execute(dpif, &op->u.execute);
ops++;
n_ops--;
}
}
} else {
size_t i;
for (i = 0; i < n_ops; i++) {
struct dpif_op *op = ops[i];
switch (op->type) {
case DPIF_OP_FLOW_PUT:
op->error = dpif_flow_put__(dpif, &op->u.flow_put);
break;
case DPIF_OP_FLOW_DEL:
op->error = dpif_flow_del__(dpif, &op->u.flow_del);
break;
case DPIF_OP_EXECUTE:
op->error = dpif_execute(dpif, &op->u.execute);
break;
default:
OVS_NOT_REACHED();
}
}
}
}
/* Returns a string that represents 'type', for use in log messages. */
const char *
dpif_upcall_type_to_string(enum dpif_upcall_type type)
{
switch (type) {
case DPIF_UC_MISS: return "miss";
case DPIF_UC_ACTION: return "action";
case DPIF_N_UC_TYPES: default: return "<unknown>";
}
}
/* Enables or disables receiving packets with dpif_recv() on 'dpif'. Returns 0
* if successful, otherwise a positive errno value.
*
* Turning packet receive off and then back on may change the Netlink PID
* assignments returned by dpif_port_get_pid(). If the client does this, it
* must update all of the flows that have OVS_ACTION_ATTR_USERSPACE actions
* using the new PID assignment. */
int
dpif_recv_set(struct dpif *dpif, bool enable)
{
int error = dpif->dpif_class->recv_set(dpif, enable);
log_operation(dpif, "recv_set", error);
return error;
}
datapath: Report kernel's flow key when passing packets up to userspace. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. This commit takes one step in that direction by making the kernel report its idea of the flow that a packet belongs to whenever it passes a packet up to userspace. This means that userspace can intelligently figure out what to do: - If userspace's notion of the flow for the packet matches the kernel's, then nothing special is necessary. - If the kernel has a more specific notion for the flow than userspace, for example if the kernel decoded IPv6 headers but userspace stopped at the Ethernet type (because it does not understand IPv6), then again nothing special is necessary: userspace can still set up the flow in the usual way. - If userspace has a more specific notion for the flow than the kernel, for example if userspace decoded an IPv6 header but the kernel stopped at the Ethernet type, then userspace can forward the packet manually, without setting up a flow in the kernel. (This case is bad from a performance point of view, but at least it is correct.) This commit does not actually make userspace flexible enough to handle changes in the kernel flow key structure, although userspace does now have enough information to do that intelligently. This will have to wait for later commits. This commit is bigger than it would otherwise be because it is rolled together with changing "struct odp_msg" to a sequence of Netlink attributes. The alternative, to do each of those changes in a separate patch, seemed like overkill because it meant that either we would have to introduce and then kill off Netlink attributes for in_port and tun_id, if Netlink conversion went first, or shove yet another variable-length header into the stuff already after odp_msg, if adding the flow key to odp_msg went first. This commit will slow down performance of checksumming packets sent up to userspace. I'm not entirely pleased with how I did it. I considered a couple of alternatives, but none of them seemed that much better. Suggestions welcome. Not changing anything wasn't an option, unfortunately. At any rate some slowdown will become unavoidable when OVS actually starts using Netlink instead of just Netlink framing. (Actually, I thought of one option where we could avoid that: make userspace do the checksum instead, by passing csum_start and csum_offset as part of what goes to userspace. But that's not perfect either.) Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2011-01-24 14:59:57 -08:00
/* Polls for an upcall from 'dpif'. If successful, stores the upcall into
* '*upcall', using 'buf' for storage. Should only be called if
* dpif_recv_set() has been used to enable receiving packets on 'dpif'.
*
* 'upcall->key' and 'upcall->userdata' point into data in the caller-provided
* 'buf', so their memory cannot be freed separately from 'buf'.
datapath: Report kernel's flow key when passing packets up to userspace. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. This commit takes one step in that direction by making the kernel report its idea of the flow that a packet belongs to whenever it passes a packet up to userspace. This means that userspace can intelligently figure out what to do: - If userspace's notion of the flow for the packet matches the kernel's, then nothing special is necessary. - If the kernel has a more specific notion for the flow than userspace, for example if the kernel decoded IPv6 headers but userspace stopped at the Ethernet type (because it does not understand IPv6), then again nothing special is necessary: userspace can still set up the flow in the usual way. - If userspace has a more specific notion for the flow than the kernel, for example if userspace decoded an IPv6 header but the kernel stopped at the Ethernet type, then userspace can forward the packet manually, without setting up a flow in the kernel. (This case is bad from a performance point of view, but at least it is correct.) This commit does not actually make userspace flexible enough to handle changes in the kernel flow key structure, although userspace does now have enough information to do that intelligently. This will have to wait for later commits. This commit is bigger than it would otherwise be because it is rolled together with changing "struct odp_msg" to a sequence of Netlink attributes. The alternative, to do each of those changes in a separate patch, seemed like overkill because it meant that either we would have to introduce and then kill off Netlink attributes for in_port and tun_id, if Netlink conversion went first, or shove yet another variable-length header into the stuff already after odp_msg, if adding the flow key to odp_msg went first. This commit will slow down performance of checksumming packets sent up to userspace. I'm not entirely pleased with how I did it. I considered a couple of alternatives, but none of them seemed that much better. Suggestions welcome. Not changing anything wasn't an option, unfortunately. At any rate some slowdown will become unavoidable when OVS actually starts using Netlink instead of just Netlink framing. (Actually, I thought of one option where we could avoid that: make userspace do the checksum instead, by passing csum_start and csum_offset as part of what goes to userspace. But that's not perfect either.) Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2011-01-24 14:59:57 -08:00
*
* The caller owns the data of 'upcall->packet' and may modify it. If
* packet's headroom is exhausted as it is manipulated, 'upcall->packet'
* will be reallocated. This requires the data of 'upcall->packet' to be
* released with ofpbuf_uninit() before 'upcall' is destroyed. However,
* when an error is returned, the 'upcall->packet' may be uninitialized
* and should not be released.
*
* Returns 0 if successful, otherwise a positive errno value. Returns EAGAIN
datapath: Report kernel's flow key when passing packets up to userspace. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. This commit takes one step in that direction by making the kernel report its idea of the flow that a packet belongs to whenever it passes a packet up to userspace. This means that userspace can intelligently figure out what to do: - If userspace's notion of the flow for the packet matches the kernel's, then nothing special is necessary. - If the kernel has a more specific notion for the flow than userspace, for example if the kernel decoded IPv6 headers but userspace stopped at the Ethernet type (because it does not understand IPv6), then again nothing special is necessary: userspace can still set up the flow in the usual way. - If userspace has a more specific notion for the flow than the kernel, for example if userspace decoded an IPv6 header but the kernel stopped at the Ethernet type, then userspace can forward the packet manually, without setting up a flow in the kernel. (This case is bad from a performance point of view, but at least it is correct.) This commit does not actually make userspace flexible enough to handle changes in the kernel flow key structure, although userspace does now have enough information to do that intelligently. This will have to wait for later commits. This commit is bigger than it would otherwise be because it is rolled together with changing "struct odp_msg" to a sequence of Netlink attributes. The alternative, to do each of those changes in a separate patch, seemed like overkill because it meant that either we would have to introduce and then kill off Netlink attributes for in_port and tun_id, if Netlink conversion went first, or shove yet another variable-length header into the stuff already after odp_msg, if adding the flow key to odp_msg went first. This commit will slow down performance of checksumming packets sent up to userspace. I'm not entirely pleased with how I did it. I considered a couple of alternatives, but none of them seemed that much better. Suggestions welcome. Not changing anything wasn't an option, unfortunately. At any rate some slowdown will become unavoidable when OVS actually starts using Netlink instead of just Netlink framing. (Actually, I thought of one option where we could avoid that: make userspace do the checksum instead, by passing csum_start and csum_offset as part of what goes to userspace. But that's not perfect either.) Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2011-01-24 14:59:57 -08:00
* if no upcall is immediately available. */
int
dpif_recv(struct dpif *dpif, struct dpif_upcall *upcall, struct ofpbuf *buf)
{
int error = dpif->dpif_class->recv(dpif, upcall, buf);
datapath: Report kernel's flow key when passing packets up to userspace. One of the goals for Open vSwitch is to decouple kernel and userspace software, so that either one can be upgraded or rolled back independent of the other. To do this in full generality, it must be possible to change the kernel's idea of the flow key separately from the userspace version. This commit takes one step in that direction by making the kernel report its idea of the flow that a packet belongs to whenever it passes a packet up to userspace. This means that userspace can intelligently figure out what to do: - If userspace's notion of the flow for the packet matches the kernel's, then nothing special is necessary. - If the kernel has a more specific notion for the flow than userspace, for example if the kernel decoded IPv6 headers but userspace stopped at the Ethernet type (because it does not understand IPv6), then again nothing special is necessary: userspace can still set up the flow in the usual way. - If userspace has a more specific notion for the flow than the kernel, for example if userspace decoded an IPv6 header but the kernel stopped at the Ethernet type, then userspace can forward the packet manually, without setting up a flow in the kernel. (This case is bad from a performance point of view, but at least it is correct.) This commit does not actually make userspace flexible enough to handle changes in the kernel flow key structure, although userspace does now have enough information to do that intelligently. This will have to wait for later commits. This commit is bigger than it would otherwise be because it is rolled together with changing "struct odp_msg" to a sequence of Netlink attributes. The alternative, to do each of those changes in a separate patch, seemed like overkill because it meant that either we would have to introduce and then kill off Netlink attributes for in_port and tun_id, if Netlink conversion went first, or shove yet another variable-length header into the stuff already after odp_msg, if adding the flow key to odp_msg went first. This commit will slow down performance of checksumming packets sent up to userspace. I'm not entirely pleased with how I did it. I considered a couple of alternatives, but none of them seemed that much better. Suggestions welcome. Not changing anything wasn't an option, unfortunately. At any rate some slowdown will become unavoidable when OVS actually starts using Netlink instead of just Netlink framing. (Actually, I thought of one option where we could avoid that: make userspace do the checksum instead, by passing csum_start and csum_offset as part of what goes to userspace. But that's not perfect either.) Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
2011-01-24 14:59:57 -08:00
if (!error && !VLOG_DROP_DBG(&dpmsg_rl)) {
struct ds flow;
char *packet;
packet = ofp_packet_to_string(upcall->packet.data,
upcall->packet.size);
ds_init(&flow);
odp_flow_key_format(upcall->key, upcall->key_len, &flow);
VLOG_DBG("%s: %s upcall:\n%s\n%s",
dpif_name(dpif), dpif_upcall_type_to_string(upcall->type),
ds_cstr(&flow), packet);
ds_destroy(&flow);
free(packet);
} else if (error && error != EAGAIN) {
log_operation(dpif, "recv", error);
}
return error;
}
/* Discards all messages that would otherwise be received by dpif_recv() on
* 'dpif'. */
void
dpif_recv_purge(struct dpif *dpif)
{
COVERAGE_INC(dpif_purge);
if (dpif->dpif_class->recv_purge) {
dpif->dpif_class->recv_purge(dpif);
}
}
/* Arranges for the poll loop to wake up when 'dpif' has a message queued to be
* received with dpif_recv(). */
void
dpif_recv_wait(struct dpif *dpif)
{
dpif->dpif_class->recv_wait(dpif);
}
/* Obtains the NetFlow engine type and engine ID for 'dpif' into '*engine_type'
* and '*engine_id', respectively. */
void
dpif_get_netflow_ids(const struct dpif *dpif,
uint8_t *engine_type, uint8_t *engine_id)
{
*engine_type = dpif->netflow_engine_type;
*engine_id = dpif->netflow_engine_id;
}
/* Translates OpenFlow queue ID 'queue_id' (in host byte order) into a priority
* value used for setting packet priority.
* On success, returns 0 and stores the priority into '*priority'.
* On failure, returns a positive errno value and stores 0 into '*priority'. */
int
dpif_queue_to_priority(const struct dpif *dpif, uint32_t queue_id,
uint32_t *priority)
{
int error = (dpif->dpif_class->queue_to_priority
? dpif->dpif_class->queue_to_priority(dpif, queue_id,
priority)
: EOPNOTSUPP);
if (error) {
*priority = 0;
}
log_operation(dpif, "queue_to_priority", error);
return error;
}
void
dpif_init(struct dpif *dpif, const struct dpif_class *dpif_class,
const char *name,
uint8_t netflow_engine_type, uint8_t netflow_engine_id)
{
dpif->dpif_class = dpif_class;
dpif->base_name = xstrdup(name);
dpif->full_name = xasprintf("%s@%s", dpif_class->type, name);
dpif->netflow_engine_type = netflow_engine_type;
dpif->netflow_engine_id = netflow_engine_id;
}
/* Undoes the results of initialization.
*
* Normally this function only needs to be called from dpif_close().
* However, it may be called by providers due to an error on opening
* that occurs after initialization. It this case dpif_close() would
* never be called. */
void
dpif_uninit(struct dpif *dpif, bool close)
{
char *base_name = dpif->base_name;
char *full_name = dpif->full_name;
if (close) {
dpif->dpif_class->close(dpif);
}
free(base_name);
free(full_name);
}
static void
log_operation(const struct dpif *dpif, const char *operation, int error)
{
if (!error) {
VLOG_DBG_RL(&dpmsg_rl, "%s: %s success", dpif_name(dpif), operation);
} else if (ofperr_is_valid(error)) {
VLOG_WARN_RL(&error_rl, "%s: %s failed (%s)",
dpif_name(dpif), operation, ofperr_get_name(error));
} else {
VLOG_WARN_RL(&error_rl, "%s: %s failed (%s)",
dpif_name(dpif), operation, ovs_strerror(error));
}
}
static enum vlog_level
flow_message_log_level(int error)
{
/* If flows arrive in a batch, userspace may push down multiple
* unique flow definitions that overlap when wildcards are applied.
* Kernels that support flow wildcarding will reject these flows as
* duplicates (EEXIST), so lower the log level to debug for these
* types of messages. */
return (error && error != EEXIST) ? VLL_WARN : VLL_DBG;
}
static bool
should_log_flow_message(int error)
{
return !vlog_should_drop(THIS_MODULE, flow_message_log_level(error),
error ? &error_rl : &dpmsg_rl);
}
static void
log_flow_message(const struct dpif *dpif, int error, const char *operation,
const struct nlattr *key, size_t key_len,
const struct nlattr *mask, size_t mask_len,
const struct dpif_flow_stats *stats,
const struct nlattr *actions, size_t actions_len)
{
struct ds ds = DS_EMPTY_INITIALIZER;
ds_put_format(&ds, "%s: ", dpif_name(dpif));
if (error) {
ds_put_cstr(&ds, "failed to ");
}
ds_put_format(&ds, "%s ", operation);
if (error) {
ds_put_format(&ds, "(%s) ", ovs_strerror(error));
}
odp_flow_format(key, key_len, mask, mask_len, NULL, &ds, true);
if (stats) {
ds_put_cstr(&ds, ", ");
dpif_flow_stats_format(stats, &ds);
}
if (actions || actions_len) {
ds_put_cstr(&ds, ", actions:");
format_odp_actions(&ds, actions, actions_len);
}
vlog(THIS_MODULE, flow_message_log_level(error), "%s", ds_cstr(&ds));
ds_destroy(&ds);
}
static void
log_flow_put_message(struct dpif *dpif, const struct dpif_flow_put *put,
int error)
{
if (should_log_flow_message(error)) {
struct ds s;
ds_init(&s);
ds_put_cstr(&s, "put");
if (put->flags & DPIF_FP_CREATE) {
ds_put_cstr(&s, "[create]");
}
if (put->flags & DPIF_FP_MODIFY) {
ds_put_cstr(&s, "[modify]");
}
if (put->flags & DPIF_FP_ZERO_STATS) {
ds_put_cstr(&s, "[zero]");
}
log_flow_message(dpif, error, ds_cstr(&s),
put->key, put->key_len, put->mask, put->mask_len,
put->stats, put->actions, put->actions_len);
ds_destroy(&s);
}
}
static void
log_flow_del_message(struct dpif *dpif, const struct dpif_flow_del *del,
int error)
{
if (should_log_flow_message(error)) {
log_flow_message(dpif, error, "flow_del", del->key, del->key_len,
NULL, 0, !error ? del->stats : NULL, NULL, 0);
}
}
static void
log_execute_message(struct dpif *dpif, const struct dpif_execute *execute,
int error)
{
if (!(error ? VLOG_DROP_WARN(&error_rl) : VLOG_DROP_DBG(&dpmsg_rl))) {
struct ds ds = DS_EMPTY_INITIALIZER;
char *packet;
packet = ofp_packet_to_string(execute->packet->data,
execute->packet->size);
ds_put_format(&ds, "%s: execute ", dpif_name(dpif));
format_odp_actions(&ds, execute->actions, execute->actions_len);
if (error) {
ds_put_format(&ds, " failed (%s)", ovs_strerror(error));
}
ds_put_format(&ds, " on packet %s", packet);
vlog(THIS_MODULE, error ? VLL_WARN : VLL_DBG, "%s", ds_cstr(&ds));
ds_destroy(&ds);
free(packet);
}
}