zephyr/subsys/net/ip/net_tc.c
Jukka Rissanen 86689030e8 net: Clarify logging in networking code
Remove network specific default and max log level setting
and start to use the zephyr logging values for those.

Remove LOG_MODULE_REGISTER() from net_core.h and place the
calls into .c files. This is done in order to avoid weird
compiler errors in some cases and to make the code look similar
as other subsystems.

Fixes #11343
Fixes #11659

Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
2018-12-07 12:00:04 +02:00

302 lines
7.6 KiB
C

/*
* Copyright (c) 2018 Intel Corporation.
*
* SPDX-License-Identifier: Apache-2.0
*/
#include <logging/log.h>
LOG_MODULE_REGISTER(net_tc, CONFIG_NET_TC_LOG_LEVEL);
#include <zephyr.h>
#include <string.h>
#include <net/net_core.h>
#include <net/net_pkt.h>
#include <net/net_stats.h>
#include "net_private.h"
#include "net_stats.h"
#include "net_tc_mapping.h"
/* Stacks for TX work queue */
NET_STACK_ARRAY_DEFINE(TX, tx_stack,
CONFIG_NET_TX_STACK_SIZE,
CONFIG_NET_TX_STACK_SIZE,
NET_TC_TX_COUNT);
/* Stacks for RX work queue */
NET_STACK_ARRAY_DEFINE(RX, rx_stack,
CONFIG_NET_RX_STACK_SIZE,
CONFIG_NET_RX_STACK_SIZE + CONFIG_NET_RX_STACK_RPL,
NET_TC_RX_COUNT);
static struct net_traffic_class tx_classes[NET_TC_TX_COUNT];
static struct net_traffic_class rx_classes[NET_TC_RX_COUNT];
void net_tc_submit_to_tx_queue(u8_t tc, struct net_pkt *pkt)
{
k_work_submit_to_queue(&tx_classes[tc].work_q, net_pkt_work(pkt));
}
void net_tc_submit_to_rx_queue(u8_t tc, struct net_pkt *pkt)
{
k_work_submit_to_queue(&rx_classes[tc].work_q, net_pkt_work(pkt));
}
int net_tx_priority2tc(enum net_priority prio)
{
if (prio > NET_PRIORITY_NC) {
/* Use default value suggested in 802.1Q */
prio = NET_PRIORITY_BE;
}
return tx_prio2tc_map[prio];
}
int net_rx_priority2tc(enum net_priority prio)
{
if (prio > NET_PRIORITY_NC) {
/* Use default value suggested in 802.1Q */
prio = NET_PRIORITY_BE;
}
return rx_prio2tc_map[prio];
}
/* Convert traffic class to thread priority */
static u8_t tx_tc2thread(u8_t tc)
{
/* Initial implementation just maps the traffic class to certain queue.
* If there are less queues than classes, then map them into
* some specific queue. In order to make this work same way as before,
* the thread priority 7 is used to map the default traffic class so
* this system works same way as before when TX thread default priority
* was 7.
*
* Lower value in this table means higher thread priority. The
* value is used as a parameter to K_PRIO_COOP() which converts it
* to actual thread priority.
*
* Higher traffic class value means higher priority queue. This means
* that thread_priorities[7] value should contain the highest priority
* for the TX queue handling thread.
*/
static const u8_t thread_priorities[] = {
#if NET_TC_TX_COUNT == 1
7
#endif
#if NET_TC_TX_COUNT == 2
8, 7
#endif
#if NET_TC_TX_COUNT == 3
8, 7, 6
#endif
#if NET_TC_TX_COUNT == 4
8, 7, 6, 5
#endif
#if NET_TC_TX_COUNT == 5
8, 7, 6, 5, 4
#endif
#if NET_TC_TX_COUNT == 6
8, 7, 6, 5, 4, 3
#endif
#if NET_TC_TX_COUNT == 7
8, 7, 6, 5, 4, 3, 2
#endif
#if NET_TC_TX_COUNT == 8
8, 7, 6, 5, 4, 3, 2, 1
#endif
};
BUILD_ASSERT_MSG(NET_TC_TX_COUNT <= CONFIG_NUM_COOP_PRIORITIES,
"Too many traffic classes");
NET_ASSERT(tc < ARRAY_SIZE(thread_priorities));
return thread_priorities[tc];
}
/* Convert traffic class to thread priority */
static u8_t rx_tc2thread(u8_t tc)
{
/* Initial implementation just maps the traffic class to certain queue.
* If there are less queues than classes, then map them into
* some specific queue. In order to make this work same way as before,
* the thread priority 7 is used to map the default traffic class so
* this system works same way as before when RX thread default priority
* was 7.
*
* Lower value in this table means higher thread priority. The
* value is used as a parameter to K_PRIO_COOP() which converts it
* to actual thread priority.
*
* Higher traffic class value means higher priority queue. This means
* that thread_priorities[7] value should contain the highest priority
* for the RX queue handling thread.
*/
static const u8_t thread_priorities[] = {
#if NET_TC_RX_COUNT == 1
7
#endif
#if NET_TC_RX_COUNT == 2
8, 7
#endif
#if NET_TC_RX_COUNT == 3
8, 7, 6
#endif
#if NET_TC_RX_COUNT == 4
8, 7, 6, 5
#endif
#if NET_TC_RX_COUNT == 5
8, 7, 6, 5, 4
#endif
#if NET_TC_RX_COUNT == 6
8, 7, 6, 5, 4, 3
#endif
#if NET_TC_RX_COUNT == 7
8, 7, 6, 5, 4, 3, 2
#endif
#if NET_TC_RX_COUNT == 8
8, 7, 6, 5, 4, 3, 2, 1
#endif
};
BUILD_ASSERT_MSG(NET_TC_RX_COUNT <= CONFIG_NUM_COOP_PRIORITIES,
"Too many traffic classes");
NET_ASSERT(tc < ARRAY_SIZE(thread_priorities));
return thread_priorities[tc];
}
#if defined(CONFIG_NET_SHELL)
#define TX_STACK(idx) NET_STACK_GET_NAME(TX, tx_stack, 0)[idx].stack
#define RX_STACK(idx) NET_STACK_GET_NAME(RX, rx_stack, 0)[idx].stack
#else
#define TX_STACK(idx) NET_STACK_GET_NAME(TX, tx_stack, 0)[idx]
#define RX_STACK(idx) NET_STACK_GET_NAME(RX, rx_stack, 0)[idx]
#endif
#if defined(CONFIG_NET_STATISTICS)
/* Fixup the traffic class statistics so that "net stats" shell command will
* print output correctly.
*/
static void tc_tx_stats_priority_setup(struct net_if *iface)
{
int i;
for (i = 0; i < 8; i++) {
net_stats_update_tc_sent_priority(iface, net_tx_priority2tc(i),
i);
}
}
static void tc_rx_stats_priority_setup(struct net_if *iface)
{
int i;
for (i = 0; i < 8; i++) {
net_stats_update_tc_recv_priority(iface, net_rx_priority2tc(i),
i);
}
}
static void net_tc_tx_stats_priority_setup(struct net_if *iface,
void *user_data)
{
ARG_UNUSED(user_data);
tc_tx_stats_priority_setup(iface);
}
static void net_tc_rx_stats_priority_setup(struct net_if *iface,
void *user_data)
{
ARG_UNUSED(user_data);
tc_rx_stats_priority_setup(iface);
}
#endif
/* Create workqueue for each traffic class we are using. All the network
* traffic goes through these classes. There needs to be at least one traffic
* class in the system.
*/
void net_tc_tx_init(void)
{
int i;
BUILD_ASSERT(NET_TC_TX_COUNT > 0);
#if defined(CONFIG_NET_STATISTICS)
net_if_foreach(net_tc_tx_stats_priority_setup, NULL);
#endif
for (i = 0; i < NET_TC_TX_COUNT; i++) {
u8_t thread_priority;
thread_priority = tx_tc2thread(i);
tx_classes[i].tc = thread_priority;
#if defined(CONFIG_NET_SHELL)
/* Fix the thread start address so that "net stacks"
* command will print correct stack information.
*/
NET_STACK_GET_NAME(TX, tx_stack, 0)[i].stack = tx_stack[i];
NET_STACK_GET_NAME(TX, tx_stack, 0)[i].prio = thread_priority;
NET_STACK_GET_NAME(TX, tx_stack, 0)[i].idx = i;
#endif
NET_DBG("[%d] Starting TX queue %p stack %p size %zd "
"prio %d (%d)", i,
&tx_classes[i].work_q.queue, TX_STACK(i),
K_THREAD_STACK_SIZEOF(tx_stack[i]),
thread_priority, K_PRIO_COOP(thread_priority));
k_work_q_start(&tx_classes[i].work_q,
tx_stack[i],
K_THREAD_STACK_SIZEOF(tx_stack[i]),
K_PRIO_COOP(thread_priority));
k_thread_name_set(&tx_classes[i].work_q.thread, "tx_workq");
}
}
void net_tc_rx_init(void)
{
int i;
BUILD_ASSERT(NET_TC_RX_COUNT > 0);
#if defined(CONFIG_NET_STATISTICS)
net_if_foreach(net_tc_rx_stats_priority_setup, NULL);
#endif
for (i = 0; i < NET_TC_RX_COUNT; i++) {
u8_t thread_priority;
thread_priority = rx_tc2thread(i);
rx_classes[i].tc = thread_priority;
#if defined(CONFIG_NET_SHELL)
/* Fix the thread start address so that "net stacks"
* command will print correct stack information.
*/
NET_STACK_GET_NAME(RX, rx_stack, 0)[i].stack = rx_stack[i];
NET_STACK_GET_NAME(RX, rx_stack, 0)[i].prio = thread_priority;
NET_STACK_GET_NAME(RX, rx_stack, 0)[i].idx = i;
#endif
NET_DBG("[%d] Starting RX queue %p stack %p size %zd "
"prio %d (%d)", i,
&rx_classes[i].work_q.queue, RX_STACK(i),
K_THREAD_STACK_SIZEOF(rx_stack[i]),
thread_priority, K_PRIO_COOP(thread_priority));
k_work_q_start(&rx_classes[i].work_q,
rx_stack[i],
K_THREAD_STACK_SIZEOF(rx_stack[i]),
K_PRIO_COOP(thread_priority));
k_thread_name_set(&rx_classes[i].work_q.thread, "rx_workq");
}
}