-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2017 Intel Corporation. All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Intel Corporation
*/
#include <stdio.h>
#include <rte_errno.h>
#include <rte_string_fns.h>
#include <rte_eal_memconfig.h>
-#include <rte_compat.h>
-#include "rte_distributor_private.h"
-#include "rte_distributor_next.h"
-#include "rte_distributor_v20.h"
+#include <rte_pause.h>
+#include <rte_tailq.h>
-TAILQ_HEAD(rte_dist_burst_list, rte_distributor_v1705);
+#include "rte_distributor.h"
+#include "rte_distributor_single.h"
+#include "distributor_private.h"
+
+TAILQ_HEAD(rte_dist_burst_list, rte_distributor);
static struct rte_tailq_elem rte_dist_burst_tailq = {
.name = "RTE_DIST_BURST",
/**** Burst Packet APIs called by workers ****/
void
-rte_distributor_request_pkt_v1705(struct rte_distributor_v1705 *d,
+rte_distributor_request_pkt(struct rte_distributor *d,
unsigned int worker_id, struct rte_mbuf **oldpkt,
unsigned int count)
{
- struct rte_distributor_buffer_v1705 *buf = &(d->bufs[worker_id]);
+ struct rte_distributor_buffer *buf = &(d->bufs[worker_id]);
unsigned int i;
volatile int64_t *retptr64;
if (unlikely(d->alg_type == RTE_DIST_ALG_SINGLE)) {
- rte_distributor_request_pkt(d->d_v20,
+ rte_distributor_request_pkt_single(d->d_single,
worker_id, oldpkt[0]);
return;
}
retptr64 = &(buf->retptr64[0]);
- /* Spin while handshake bits are set (scheduler clears it) */
- while (unlikely(*retptr64 & RTE_DISTRIB_GET_BUF)) {
+ /* Spin while handshake bits are set (scheduler clears it).
+ * Sync with worker on GET_BUF flag.
+ */
+ while (unlikely(__atomic_load_n(retptr64, __ATOMIC_ACQUIRE)
+ & RTE_DISTRIB_GET_BUF)) {
rte_pause();
uint64_t t = rte_rdtsc()+100;
/*
* Finally, set the GET_BUF to signal to distributor that cache
* line is ready for processing
+ * Sync with distributor to release retptrs
*/
- *retptr64 |= RTE_DISTRIB_GET_BUF;
+ __atomic_store_n(retptr64, *retptr64 | RTE_DISTRIB_GET_BUF,
+ __ATOMIC_RELEASE);
}
int
-rte_distributor_poll_pkt_v1705(struct rte_distributor_v1705 *d,
+rte_distributor_poll_pkt(struct rte_distributor *d,
unsigned int worker_id, struct rte_mbuf **pkts)
{
- struct rte_distributor_buffer_v1705 *buf = &d->bufs[worker_id];
+ struct rte_distributor_buffer *buf = &d->bufs[worker_id];
uint64_t ret;
int count = 0;
unsigned int i;
if (unlikely(d->alg_type == RTE_DIST_ALG_SINGLE)) {
- pkts[0] = rte_distributor_poll_pkt(d->d_v20, worker_id);
+ pkts[0] = rte_distributor_poll_pkt_single(d->d_single,
+ worker_id);
return (pkts[0]) ? 1 : 0;
}
- /* If bit is set, return */
- if (buf->bufptr64[0] & RTE_DISTRIB_GET_BUF)
+ /* If bit is set, return
+ * Sync with distributor to acquire bufptrs
+ */
+ if (__atomic_load_n(&(buf->bufptr64[0]), __ATOMIC_ACQUIRE)
+ & RTE_DISTRIB_GET_BUF)
return -1;
/* since bufptr64 is signed, this should be an arithmetic shift */
* so now we've got the contents of the cacheline into an array of
* mbuf pointers, so toggle the bit so scheduler can start working
* on the next cacheline while we're working.
+ * Sync with distributor on GET_BUF flag. Release bufptrs.
*/
- buf->bufptr64[0] |= RTE_DISTRIB_GET_BUF;
+ __atomic_store_n(&(buf->bufptr64[0]),
+ buf->bufptr64[0] | RTE_DISTRIB_GET_BUF, __ATOMIC_RELEASE);
return count;
}
int
-rte_distributor_get_pkt_v1705(struct rte_distributor_v1705 *d,
+rte_distributor_get_pkt(struct rte_distributor *d,
unsigned int worker_id, struct rte_mbuf **pkts,
struct rte_mbuf **oldpkt, unsigned int return_count)
{
if (unlikely(d->alg_type == RTE_DIST_ALG_SINGLE)) {
if (return_count <= 1) {
- pkts[0] = rte_distributor_get_pkt(d->d_v20,
+ pkts[0] = rte_distributor_get_pkt_single(d->d_single,
worker_id, oldpkt[0]);
return (pkts[0]) ? 1 : 0;
} else
return -EINVAL;
}
- rte_distributor_request_pkt_v1705(d, worker_id, oldpkt, return_count);
+ rte_distributor_request_pkt(d, worker_id, oldpkt, return_count);
- count = rte_distributor_poll_pkt_v1705(d, worker_id, pkts);
+ count = rte_distributor_poll_pkt(d, worker_id, pkts);
while (count == -1) {
uint64_t t = rte_rdtsc() + 100;
while (rte_rdtsc() < t)
rte_pause();
- count = rte_distributor_poll_pkt_v1705(d, worker_id, pkts);
+ count = rte_distributor_poll_pkt(d, worker_id, pkts);
}
return count;
}
int
-rte_distributor_return_pkt_v1705(struct rte_distributor_v1705 *d,
+rte_distributor_return_pkt(struct rte_distributor *d,
unsigned int worker_id, struct rte_mbuf **oldpkt, int num)
{
- struct rte_distributor_buffer_v1705 *buf = &d->bufs[worker_id];
+ struct rte_distributor_buffer *buf = &d->bufs[worker_id];
unsigned int i;
if (unlikely(d->alg_type == RTE_DIST_ALG_SINGLE)) {
if (num == 1)
- return rte_distributor_return_pkt(d->d_v20,
+ return rte_distributor_return_pkt_single(d->d_single,
worker_id, oldpkt[0]);
else
return -EINVAL;
}
+ /* Sync with distributor to acquire retptrs */
+ __atomic_thread_fence(__ATOMIC_ACQUIRE);
for (i = 0; i < RTE_DIST_BURST_SIZE; i++)
/* Switch off the return bit first */
buf->retptr64[i] &= ~RTE_DISTRIB_RETURN_BUF;
buf->retptr64[i] = (((int64_t)(uintptr_t)oldpkt[i]) <<
RTE_DISTRIB_FLAG_BITS) | RTE_DISTRIB_RETURN_BUF;
- /* set the GET_BUF but even if we got no returns */
- buf->retptr64[0] |= RTE_DISTRIB_GET_BUF;
+ /* set the GET_BUF but even if we got no returns.
+ * Sync with distributor on GET_BUF flag. Release retptrs.
+ */
+ __atomic_store_n(&(buf->retptr64[0]),
+ buf->retptr64[0] | RTE_DISTRIB_GET_BUF, __ATOMIC_RELEASE);
return 0;
}
/* stores a packet returned from a worker inside the returns array */
static inline void
-store_return(uintptr_t oldbuf, struct rte_distributor_v1705 *d,
+store_return(uintptr_t oldbuf, struct rte_distributor *d,
unsigned int *ret_start, unsigned int *ret_count)
{
if (!oldbuf)
* workers to give us our atomic flow pinning.
*/
void
-find_match_scalar(struct rte_distributor_v1705 *d,
+find_match_scalar(struct rte_distributor *d,
uint16_t *data_ptr,
uint16_t *output_ptr)
{
* the valid returned pointers (store_return).
*/
static unsigned int
-handle_returns(struct rte_distributor_v1705 *d, unsigned int wkr)
+handle_returns(struct rte_distributor *d, unsigned int wkr)
{
- struct rte_distributor_buffer_v1705 *buf = &(d->bufs[wkr]);
+ struct rte_distributor_buffer *buf = &(d->bufs[wkr]);
uintptr_t oldbuf;
unsigned int ret_start = d->returns.start,
ret_count = d->returns.count;
unsigned int count = 0;
unsigned int i;
- if (buf->retptr64[0] & RTE_DISTRIB_GET_BUF) {
+ /* Sync on GET_BUF flag. Acquire retptrs. */
+ if (__atomic_load_n(&(buf->retptr64[0]), __ATOMIC_ACQUIRE)
+ & RTE_DISTRIB_GET_BUF) {
for (i = 0; i < RTE_DIST_BURST_SIZE; i++) {
if (buf->retptr64[i] & RTE_DISTRIB_RETURN_BUF) {
oldbuf = ((uintptr_t)(buf->retptr64[i] >>
}
d->returns.start = ret_start;
d->returns.count = ret_count;
- /* Clear for the worker to populate with more returns */
- buf->retptr64[0] = 0;
+ /* Clear for the worker to populate with more returns.
+ * Sync with distributor on GET_BUF flag. Release retptrs.
+ */
+ __atomic_store_n(&(buf->retptr64[0]), 0, __ATOMIC_RELEASE);
}
return count;
}
* before sending out new packets.
*/
static unsigned int
-release(struct rte_distributor_v1705 *d, unsigned int wkr)
+release(struct rte_distributor *d, unsigned int wkr)
{
- struct rte_distributor_buffer_v1705 *buf = &(d->bufs[wkr]);
+ struct rte_distributor_buffer *buf = &(d->bufs[wkr]);
unsigned int i;
- while (!(d->bufs[wkr].bufptr64[0] & RTE_DISTRIB_GET_BUF))
+ /* Sync with worker on GET_BUF flag */
+ while (!(__atomic_load_n(&(d->bufs[wkr].bufptr64[0]), __ATOMIC_ACQUIRE)
+ & RTE_DISTRIB_GET_BUF))
rte_pause();
handle_returns(d, wkr);
d->backlog[wkr].count = 0;
- /* Clear the GET bit */
- buf->bufptr64[0] &= ~RTE_DISTRIB_GET_BUF;
+ /* Clear the GET bit.
+ * Sync with worker on GET_BUF flag. Release bufptrs.
+ */
+ __atomic_store_n(&(buf->bufptr64[0]),
+ buf->bufptr64[0] & ~RTE_DISTRIB_GET_BUF, __ATOMIC_RELEASE);
return buf->count;
}
/* process a set of packets to distribute them to workers */
int
-rte_distributor_process_v1705(struct rte_distributor_v1705 *d,
+rte_distributor_process(struct rte_distributor *d,
struct rte_mbuf **mbufs, unsigned int num_mbufs)
{
unsigned int next_idx = 0;
if (d->alg_type == RTE_DIST_ALG_SINGLE) {
/* Call the old API */
- return rte_distributor_process(d->d_v20, mbufs, num_mbufs);
+ return rte_distributor_process_single(d->d_single,
+ mbufs, num_mbufs);
}
if (unlikely(num_mbufs == 0)) {
/* Flush out all non-full cache-lines to workers. */
for (wid = 0 ; wid < d->num_workers; wid++) {
- if ((d->bufs[wid].bufptr64[0] & RTE_DISTRIB_GET_BUF)) {
+ /* Sync with worker on GET_BUF flag. */
+ if (__atomic_load_n(&(d->bufs[wid].bufptr64[0]),
+ __ATOMIC_ACQUIRE) & RTE_DISTRIB_GET_BUF) {
release(d, wid);
handle_returns(d, wid);
}
uint16_t matches[RTE_DIST_BURST_SIZE];
unsigned int pkts;
- if (d->bufs[wkr].bufptr64[0] & RTE_DISTRIB_GET_BUF)
+ /* Sync with worker on GET_BUF flag. */
+ if (__atomic_load_n(&(d->bufs[wkr].bufptr64[0]),
+ __ATOMIC_ACQUIRE) & RTE_DISTRIB_GET_BUF)
d->bufs[wkr].count = 0;
if ((num_mbufs - next_idx) < RTE_DIST_BURST_SIZE)
next_value = (((int64_t)(uintptr_t)next_mb) <<
RTE_DISTRIB_FLAG_BITS);
/*
- * User is advocated to set tag vaue for each
+ * User is advocated to set tag value for each
* mbuf before calling rte_distributor_process.
* User defined tags are used to identify flows,
* or sessions.
/*
* Uncommenting the next line will cause the find_match
- * function to be optimised out, making this function
+ * function to be optimized out, making this function
* do parallel (non-atomic) distribution
*/
/* matches[j] = 0; */
/* Flush out all non-full cache-lines to workers. */
for (wid = 0 ; wid < d->num_workers; wid++)
- if ((d->bufs[wid].bufptr64[0] & RTE_DISTRIB_GET_BUF))
+ /* Sync with worker on GET_BUF flag. */
+ if ((__atomic_load_n(&(d->bufs[wid].bufptr64[0]),
+ __ATOMIC_ACQUIRE) & RTE_DISTRIB_GET_BUF))
release(d, wid);
return num_mbufs;
/* return to the caller, packets returned from workers */
int
-rte_distributor_returned_pkts_v1705(struct rte_distributor_v1705 *d,
+rte_distributor_returned_pkts(struct rte_distributor *d,
struct rte_mbuf **mbufs, unsigned int max_mbufs)
{
struct rte_distributor_returned_pkts *returns = &d->returns;
if (d->alg_type == RTE_DIST_ALG_SINGLE) {
/* Call the old API */
- return rte_distributor_returned_pkts(d->d_v20,
+ return rte_distributor_returned_pkts_single(d->d_single,
mbufs, max_mbufs);
}
/*
* Return the number of packets in-flight in a distributor, i.e. packets
- * being workered on or queued up in a backlog.
+ * being worked on or queued up in a backlog.
*/
static inline unsigned int
-total_outstanding(const struct rte_distributor_v1705 *d)
+total_outstanding(const struct rte_distributor *d)
{
unsigned int wkr, total_outstanding = 0;
* queued up.
*/
int
-rte_distributor_flush_v1705(struct rte_distributor_v1705 *d)
+rte_distributor_flush(struct rte_distributor *d)
{
- const unsigned int flushed = total_outstanding(d);
+ unsigned int flushed;
unsigned int wkr;
if (d->alg_type == RTE_DIST_ALG_SINGLE) {
/* Call the old API */
- return rte_distributor_flush(d->d_v20);
+ return rte_distributor_flush_single(d->d_single);
}
+ flushed = total_outstanding(d);
+
while (total_outstanding(d) > 0)
- rte_distributor_process_v1705(d, NULL, 0);
+ rte_distributor_process(d, NULL, 0);
+
+ /* wait 10ms to allow all worker drain the pkts */
+ rte_delay_us(10000);
/*
* Send empty burst to all workers to allow them to exit
* gracefully, should they need to.
*/
- rte_distributor_process_v1705(d, NULL, 0);
+ rte_distributor_process(d, NULL, 0);
for (wkr = 0; wkr < d->num_workers; wkr++)
handle_returns(d, wkr);
/* clears the internal returns array in the distributor */
void
-rte_distributor_clear_returns_v1705(struct rte_distributor_v1705 *d)
+rte_distributor_clear_returns(struct rte_distributor *d)
{
unsigned int wkr;
if (d->alg_type == RTE_DIST_ALG_SINGLE) {
/* Call the old API */
- rte_distributor_clear_returns(d->d_v20);
+ rte_distributor_clear_returns_single(d->d_single);
+ return;
}
/* throw away returns, so workers can exit */
for (wkr = 0; wkr < d->num_workers; wkr++)
- d->bufs[wkr].retptr64[0] = 0;
+ /* Sync with worker. Release retptrs. */
+ __atomic_store_n(&(d->bufs[wkr].retptr64[0]), 0,
+ __ATOMIC_RELEASE);
}
/* creates a distributor instance */
-struct rte_distributor_v1705 *
-rte_distributor_create_v1705(const char *name,
+struct rte_distributor *
+rte_distributor_create(const char *name,
unsigned int socket_id,
unsigned int num_workers,
unsigned int alg_type)
{
- struct rte_distributor_v1705 *d;
+ struct rte_distributor *d;
struct rte_dist_burst_list *dist_burst_list;
char mz_name[RTE_MEMZONE_NAMESIZE];
const struct rte_memzone *mz;
RTE_BUILD_BUG_ON((sizeof(*d) & RTE_CACHE_LINE_MASK) != 0);
RTE_BUILD_BUG_ON((RTE_DISTRIB_MAX_WORKERS & 7) != 0);
+ if (name == NULL || num_workers >=
+ (unsigned int)RTE_MIN(RTE_DISTRIB_MAX_WORKERS, RTE_MAX_LCORE)) {
+ rte_errno = EINVAL;
+ return NULL;
+ }
+
if (alg_type == RTE_DIST_ALG_SINGLE) {
- d = malloc(sizeof(struct rte_distributor_v1705));
- d->d_v20 = rte_distributor_create(name,
+ d = malloc(sizeof(struct rte_distributor));
+ if (d == NULL) {
+ rte_errno = ENOMEM;
+ return NULL;
+ }
+ d->d_single = rte_distributor_create_single(name,
socket_id, num_workers);
- if (d->d_v20 == NULL) {
+ if (d->d_single == NULL) {
+ free(d);
/* rte_errno will have been set */
return NULL;
}
return d;
}
- if (name == NULL || num_workers >= RTE_DISTRIB_MAX_WORKERS) {
- rte_errno = EINVAL;
- return NULL;
- }
-
snprintf(mz_name, sizeof(mz_name), RTE_DISTRIB_PREFIX"%s", name);
mz = rte_memzone_reserve(mz_name, sizeof(*d), socket_id, NO_FLAGS);
if (mz == NULL) {
}
d = mz->addr;
- snprintf(d->name, sizeof(d->name), "%s", name);
+ strlcpy(d->name, name, sizeof(d->name));
d->num_workers = num_workers;
d->alg_type = alg_type;
+ d->dist_match_fn = RTE_DIST_MATCH_SCALAR;
#if defined(RTE_ARCH_X86)
- if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE4_2))
- d->dist_match_fn = RTE_DIST_MATCH_VECTOR;
- else
+ d->dist_match_fn = RTE_DIST_MATCH_VECTOR;
#endif
- d->dist_match_fn = RTE_DIST_MATCH_SCALAR;
/*
- * Set up the backog tags so they're pointing at the second cache
+ * Set up the backlog tags so they're pointing at the second cache
* line for performance during flow matching
*/
for (i = 0 ; i < num_workers ; i++)
rte_dist_burst_list);
- rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
+ rte_mcfg_tailq_write_lock();
TAILQ_INSERT_TAIL(dist_burst_list, d, next);
- rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
+ rte_mcfg_tailq_write_unlock();
return d;
}
git.droids-corp.org / dpdk.git / blobdiff