[dpdk-dev] [PATCH v6 8/8] l3fwd-power: enable one-shot rx interrupt and polling/interrupt mode switch

[Stephen Hemminger]

On Fri, 27 Feb 2015 12:56:16 +0800
Cunming Liang  wrote:

> +/* ethernet addresses of ports */
> +static rte_spinlock_t locks[RTE_MAX_ETHPORTS];

Comment is incorrect this is a lock array not an address array.


>  static struct rte_eth_conf port_conf = {
>   .rxmode = {
> - .mq_mode= ETH_MQ_RX_RSS,
> + .mq_mode = ETH_MQ_RX_RSS,

Please don't mix white space changes with code changes

[dpdk-dev] [PATCH v6 8/8] l3fwd-power: enable one-shot rx interrupt and polling/interrupt mode switch

[Stephen Hemminger]

On Fri, 27 Feb 2015 12:56:16 +0800
Cunming Liang  wrote:

> + /* Enable one-shot rx interrupt */
> + rte_spinlock_lock(&(locks[port_id]));
> + rte_eth_dev_rx_intr_enable(port_id, queue_id);
> + rte_spinlock_unlock(&(locks[port_id]));
> +

If always requires locks like this, then the API should
do the locking internally.

[dpdk-dev] [PATCH v6 8/8] l3fwd-power: enable one-shot rx interrupt and polling/interrupt mode switch

[Cunming Liang]

From: "Zhou, Danny" 

Demonstrate how to handle per rx queue interrupt in a NAPI-like
implementation in usersapce. PDK polling thread mainly works in
polling mode and switch to interrupt mode only if there is no
any packet received in recent polls.
Usersapce interrupt notification generally takes a lot more cycles
than kernel, so one-shot interrupt is used here to guarantee minimum
overhead and DPDK polling thread returns to polling mode immediately
once it receives an interrupt notificaiton for incoming packet.

Signed-off-by: Danny Zhou 
Signed-off-by: Cunming Liang 
---
v6 changes
 - Split event fd add and wait

v5 changes
 - Change invoked function name and parameter to accomodate EAL change

v3 changes
 - Add spinlock to ensure thread safe when accessing interrupt mask
   register

v2 changes
 - Remove unused function which is for debug purpose

 examples/l3fwd-power/main.c | 194 ++--
 1 file changed, 152 insertions(+), 42 deletions(-)

diff --git a/examples/l3fwd-power/main.c b/examples/l3fwd-power/main.c
index f6b55b9..9a920f2 100644
--- a/examples/l3fwd-power/main.c
+++ b/examples/l3fwd-power/main.c
@@ -75,12 +75,14 @@
 #include 
 #include 
 #include 
+#include 
+#include 

 #define RTE_LOGTYPE_L3FWD_POWER RTE_LOGTYPE_USER1

 #define MAX_PKT_BURST 32

-#define MIN_ZERO_POLL_COUNT 5
+#define MIN_ZERO_POLL_COUNT 10

 /* around 100ms at 2 Ghz */
 #define TIMER_RESOLUTION_CYCLES   2ULL
@@ -156,6 +158,9 @@ static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
 /* ethernet addresses of ports */
 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];

+/* ethernet addresses of ports */
+static rte_spinlock_t locks[RTE_MAX_ETHPORTS];
+
 /* mask of enabled ports */
 static uint32_t enabled_port_mask = 0;
 /* Ports set in promiscuous mode off by default. */
@@ -188,6 +193,9 @@ struct lcore_rx_queue {
 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
 #define MAX_RX_QUEUE_PER_PORT 128

+#define MAX_RX_QUEUE_INTERRUPT_PER_PORT 16
+
+
 #define MAX_LCORE_PARAMS 1024
 struct lcore_params {
uint8_t port_id;
@@ -214,7 +222,7 @@ static uint16_t nb_lcore_params = 
sizeof(lcore_params_array_default) /

 static struct rte_eth_conf port_conf = {
.rxmode = {
-   .mq_mode= ETH_MQ_RX_RSS,
+   .mq_mode = ETH_MQ_RX_RSS,
.max_rx_pkt_len = ETHER_MAX_LEN,
.split_hdr_size = 0,
.header_split   = 0, /**< Header Split disabled */
@@ -226,11 +234,14 @@ static struct rte_eth_conf port_conf = {
.rx_adv_conf = {
.rss_conf = {
.rss_key = NULL,
-   .rss_hf = ETH_RSS_IP,
+   .rss_hf = ETH_RSS_UDP,
},
},
.txmode = {
-   .mq_mode = ETH_DCB_NONE,
+   .mq_mode = ETH_MQ_TX_NONE,
+   },
+   .intr_conf = {
+   .rxq = 1, /**< rxq interrupt feature enabled */
},
 };

@@ -402,19 +413,22 @@ power_timer_cb(__attribute__((unused)) struct rte_timer 
*tim,
/* accumulate total execution time in us when callback is invoked */
sleep_time_ratio = (float)(stats[lcore_id].sleep_time) /
(float)SCALING_PERIOD;
-
/**
 * check whether need to scale down frequency a step if it sleep a lot.
 */
-   if (sleep_time_ratio >= SCALING_DOWN_TIME_RATIO_THRESHOLD)
-   rte_power_freq_down(lcore_id);
+   if (sleep_time_ratio >= SCALING_DOWN_TIME_RATIO_THRESHOLD) {
+   if (rte_power_freq_down)
+   rte_power_freq_down(lcore_id);
+   }
else if ( (unsigned)(stats[lcore_id].nb_rx_processed /
-   stats[lcore_id].nb_iteration_looped) < MAX_PKT_BURST)
+   stats[lcore_id].nb_iteration_looped) < MAX_PKT_BURST) {
/**
 * scale down a step if average packet per iteration less
 * than expectation.
 */
-   rte_power_freq_down(lcore_id);
+   if (rte_power_freq_down)
+   rte_power_freq_down(lcore_id);
+   }

/**
 * initialize another timer according to current frequency to ensure
@@ -707,22 +721,20 @@ l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid,

 }

-#define SLEEP_GEAR1_THRESHOLD100
-#define SLEEP_GEAR2_THRESHOLD1000
+#define MINIMUM_SLEEP_TIME 1
+#define SUSPEND_THRESHOLD  300

 static inline uint32_t
 power_idle_heuristic(uint32_t zero_rx_packet_count)
 {
-   /* If zero count is less than 100, use it as the sleep time in us */
-   if (zero_rx_packet_count < SLEEP_GEAR1_THRESHOLD)
-   return zero_rx_packet_count;
-   /* If zero count is less than 1000, sleep time should be 100 us */
-   else if ((zero_rx_packet_count >= SLEEP_GEAR1_THRESHOLD) &&
-