qcacmn: Fix tx completion and rx stats per wbm/reo ring

Tx completion stats should be counted per wbm ring on tx completion
and not globally. Similarly, rx stats should be counter per reo ring.
Change-Id: I1e4af0d38b23e60de78ca03316861db08ff0811a
This commit is contained in:
Varsha Mishra
2019-05-31 00:59:15 +05:30
committed by nshrivas
parent 3c91fb553a
commit 1f4cfb6c58
8 changed files with 238 additions and 116 deletions

View File

@@ -71,7 +71,8 @@
#define WME_AC_VO 3 /* voice */
#define WME_AC_MAX 4 /* MAX AC Value */
#define CDP_MAX_RX_RINGS 4
#define CDP_MAX_RX_RINGS 4 /* max rx rings */
#define CDP_MAX_TX_COMP_RINGS 3 /* max tx completion rings */
/* TID level VoW stats macros
* to add and get stats
@@ -291,8 +292,7 @@ struct cdp_delay_stats {
* @hwtx_delay: delay between wifi driver exit (enqueue to HW) and tx completion
* @intfrm_delay: interframe delay
* @success_cnt: total successful transmit count
* @complete_cnt: total transmit count
* @fwdrop_cnt: firmware drop found in tx completion path
* @comp_fail_cnt: firmware drop found in tx completion path
* @swdrop_cnt: software drop in tx path
*/
struct cdp_tid_tx_stats {
@@ -300,7 +300,6 @@ struct cdp_tid_tx_stats {
struct cdp_delay_stats hwtx_delay;
struct cdp_delay_stats intfrm_delay;
uint64_t success_cnt;
uint64_t complete_cnt;
uint64_t comp_fail_cnt;
uint64_t swdrop_cnt[TX_MAX_DROP];
};
@@ -330,14 +329,17 @@ struct cdp_tid_rx_stats {
/*
* struct cdp_tid_stats
* @ingress_stack: Total packets received from linux stack
* @osif_drop: drops in osif layer
* @tid_tx_stats: transmit counters per tid
* @tid_rx_stats: receive counters per tid
*/
struct cdp_tid_stats {
uint64_t ingress_stack;
uint64_t osif_drop;
struct cdp_tid_tx_stats tid_tx_stats[CDP_MAX_DATA_TIDS];
struct cdp_tid_rx_stats tid_rx_stats[CDP_MAX_DATA_TIDS];
struct cdp_tid_tx_stats tid_tx_stats[CDP_MAX_TX_COMP_RINGS]
[CDP_MAX_DATA_TIDS];
struct cdp_tid_rx_stats tid_rx_stats[CDP_MAX_RX_RINGS]
[CDP_MAX_DATA_TIDS];
};
/* struct cdp_pkt_info - packet info

View File

@@ -858,8 +858,19 @@ uint32_t dp_pdev_tid_stats_display(void *pdev_handle,
enum _ol_ath_param_t param, uint32_t value, void *buff);
#endif
/**
* dp_update_delay_stats() - Update delay statistics in structure
* and fill min, max and avg delay
* @pdev: pdev handle
* @delay: delay in ms
* @tid: tid value
* @mode: type of tx delay mode
* @ring id: ring number
*
* Return: none
*/
void dp_update_delay_stats(struct dp_pdev *pdev, uint32_t delay,
uint8_t tid, uint8_t mode);
uint8_t tid, uint8_t mode, uint8_t ring_id);
/**
* dp_print_ring_stats(): Print tail and head pointer

View File

@@ -1619,29 +1619,22 @@ static qdf_nbuf_t dp_ipa_intrabss_send(struct dp_pdev *pdev,
struct dp_vdev *vdev,
qdf_nbuf_t nbuf)
{
struct cdp_tid_rx_stats *tid_stats;
struct dp_peer *vdev_peer;
uint16_t len;
uint8_t tid;
vdev_peer = vdev->vap_bss_peer;
if (qdf_unlikely(!vdev_peer))
return nbuf;
tid = qdf_nbuf_get_priority(nbuf);
tid_stats = &pdev->stats.tid_stats.tid_rx_stats[tid];
qdf_mem_zero(nbuf->cb, sizeof(nbuf->cb));
len = qdf_nbuf_len(nbuf);
if (dp_tx_send(vdev, nbuf)) {
DP_STATS_INC_PKT(vdev_peer, rx.intra_bss.fail, 1, len);
tid_stats->fail_cnt[INTRABSS_DROP]++;
return nbuf;
}
DP_STATS_INC_PKT(vdev_peer, rx.intra_bss.pkts, 1, len);
tid_stats->intrabss_cnt++;
return NULL;
}

View File

@@ -1394,7 +1394,7 @@ static uint32_t dp_service_srngs(void *dp_ctx, uint32_t dp_budget)
work_done = dp_tx_comp_handler(int_ctx,
soc,
soc->tx_comp_ring[ring].hal_srng,
remaining_quota);
ring, remaining_quota);
if (work_done) {
intr_stats->num_tx_ring_masks[ring]++;
@@ -9635,19 +9635,20 @@ static uint8_t dp_bucket_index(uint32_t delay, uint16_t *array)
*
* @pdev: pdev handle
* @delay: delay in ms
* @t: tid value
* @tid: tid value
* @mode: type of tx delay mode
* @ring_id: ring number
* Return: pointer to cdp_delay_stats structure
*/
static struct cdp_delay_stats *
dp_fill_delay_buckets(struct dp_pdev *pdev, uint32_t delay,
uint8_t tid, uint8_t mode)
uint8_t tid, uint8_t mode, uint8_t ring_id)
{
uint8_t delay_index = 0;
struct cdp_tid_tx_stats *tstats =
&pdev->stats.tid_stats.tid_tx_stats[tid];
&pdev->stats.tid_stats.tid_tx_stats[ring_id][tid];
struct cdp_tid_rx_stats *rstats =
&pdev->stats.tid_stats.tid_rx_stats[tid];
&pdev->stats.tid_stats.tid_rx_stats[ring_id][tid];
/*
* cdp_fw_to_hw_delay_range
* Fw to hw delay ranges in milliseconds
@@ -9723,10 +9724,11 @@ dp_fill_delay_buckets(struct dp_pdev *pdev, uint32_t delay,
* @delay: delay in ms
* @tid: tid value
* @mode: type of tx delay mode
* @ring id: ring number
* Return: none
*/
void dp_update_delay_stats(struct dp_pdev *pdev, uint32_t delay,
uint8_t tid, uint8_t mode)
uint8_t tid, uint8_t mode, uint8_t ring_id)
{
struct cdp_delay_stats *dstats = NULL;
@@ -9734,7 +9736,7 @@ void dp_update_delay_stats(struct dp_pdev *pdev, uint32_t delay,
* Delay ranges are different for different delay modes
* Get the correct index to update delay bucket
*/
dstats = dp_fill_delay_buckets(pdev, delay, tid, mode);
dstats = dp_fill_delay_buckets(pdev, delay, tid, mode, ring_id);
if (qdf_unlikely(!dstats))
return;

View File

@@ -393,8 +393,9 @@ dp_rx_intrabss_fwd(struct dp_soc *soc,
struct dp_ast_entry *ast_entry;
qdf_nbuf_t nbuf_copy;
uint8_t tid = qdf_nbuf_get_tid_val(nbuf);
struct cdp_tid_rx_stats *tid_stats =
&ta_peer->vdev->pdev->stats.tid_stats.tid_rx_stats[tid];
uint8_t ring_id = QDF_NBUF_CB_RX_CTX_ID(nbuf);
struct cdp_tid_rx_stats *tid_stats = &ta_peer->vdev->pdev->stats.
tid_stats.tid_rx_stats[ring_id][tid];
/* check if the destination peer is available in peer table
* and also check if the source peer and destination peer
@@ -1078,6 +1079,7 @@ qdf_nbuf_t dp_rx_sg_create(qdf_nbuf_t nbuf, uint8_t *rx_tlv_hdr)
*/
void dp_rx_compute_delay(struct dp_vdev *vdev, qdf_nbuf_t nbuf)
{
uint8_t ring_id = QDF_NBUF_CB_RX_CTX_ID(nbuf);
int64_t current_ts = qdf_ktime_to_ms(qdf_ktime_get());
uint32_t to_stack = qdf_nbuf_get_timedelta_ms(nbuf);
uint8_t tid = qdf_nbuf_get_tid_val(nbuf);
@@ -1085,7 +1087,7 @@ void dp_rx_compute_delay(struct dp_vdev *vdev, qdf_nbuf_t nbuf)
(uint32_t)(current_ts - vdev->prev_rx_deliver_tstamp);
dp_update_delay_stats(vdev->pdev, to_stack, tid,
CDP_DELAY_STATS_REAP_STACK);
CDP_DELAY_STATS_REAP_STACK, ring_id);
/*
* Update interframe delay stats calculated at deliver_data_ol point.
* Value of vdev->prev_rx_deliver_tstamp will be 0 for 1st frame, so
@@ -1094,7 +1096,7 @@ void dp_rx_compute_delay(struct dp_vdev *vdev, qdf_nbuf_t nbuf)
* of vdev->prev_rx_deliver_tstamp.
*/
dp_update_delay_stats(vdev->pdev, interframe_delay, tid,
CDP_DELAY_STATS_RX_INTERFRAME);
CDP_DELAY_STATS_RX_INTERFRAME, ring_id);
vdev->prev_rx_deliver_tstamp = current_ts;
}
@@ -1109,15 +1111,16 @@ static inline int dp_rx_drop_nbuf_list(struct dp_pdev *pdev,
qdf_nbuf_t buf_list)
{
struct cdp_tid_rx_stats *stats = NULL;
uint8_t tid = 0;
uint8_t tid = 0, ring_id = 0;
int num_dropped = 0;
qdf_nbuf_t buf, next_buf;
buf = buf_list;
while (buf) {
ring_id = QDF_NBUF_CB_RX_CTX_ID(buf);
next_buf = qdf_nbuf_queue_next(buf);
tid = qdf_nbuf_get_tid_val(buf);
stats = &pdev->stats.tid_stats.tid_rx_stats[tid];
stats = &pdev->stats.tid_stats.tid_rx_stats[ring_id][tid];
stats->fail_cnt[INVALID_PEER_VDEV]++;
stats->delivered_to_stack--;
qdf_nbuf_free(buf);
@@ -1816,7 +1819,8 @@ done:
if (qdf_unlikely(rx_pdev->delay_stats_flag))
qdf_nbuf_set_timestamp(nbuf);
tid_stats = &rx_pdev->stats.tid_stats.tid_rx_stats[tid];
tid_stats = &rx_pdev->stats.tid_stats.
tid_rx_stats[ring_id][tid];
if (qdf_unlikely(!hal_rx_attn_msdu_done_get(rx_tlv_hdr))) {
dp_err("MSDU DONE failure");
DP_STATS_INC(soc, rx.err.msdu_done_fail, 1);

View File

@@ -186,6 +186,9 @@ const char *intfrm_delay_bucket[CDP_DELAY_BUCKET_MAX + 1] = {
#include "dp_tx_capture.h"
#endif
#define TID_COUNTER_STATS 1 /* Success/drop stats type */
#define TID_DELAY_STATS 2 /* Delay stats type */
/*
* dp_print_stats_string_tlv: display htt_stats_string_tlv
* @tag_buf: buffer containing the tlv htt_stats_string_tlv
@@ -4030,79 +4033,174 @@ static inline const char *dp_vow_str_intfrm_delay(uint8_t index)
return intfrm_delay_bucket[index];
}
/**
* dp_accumulate_delay_stats() - Update delay stats members
* @total: Update stats total structure
* @per_ring: per ring structures from where stats need to be accumulated
*
* Return: void
*/
static void
dp_accumulate_delay_stats(struct cdp_delay_stats *total,
struct cdp_delay_stats *per_ring)
{
uint8_t index;
for (index = 0; index < CDP_DELAY_BUCKET_MAX; index++)
total->delay_bucket[index] += per_ring->delay_bucket[index];
total->min_delay = QDF_MIN(total->min_delay, per_ring->min_delay);
total->max_delay = QDF_MAX(total->max_delay, per_ring->max_delay);
total->avg_delay = (total->avg_delay + per_ring->avg_delay) / 2;
}
/**
* dp_accumulate_tid_stats() - Accumulate TID stats from each ring
* @pdev: pdev handle
* @tid: traffic ID
* @total_tx: fill this tx structure to get stats from all wbm rings
* @total_rx: fill this rx structure to get stats from all reo rings
* @type: delay stats or regular frame counters
*
* Return: void
*/
static void
dp_accumulate_tid_stats(struct dp_pdev *pdev, uint8_t tid,
struct cdp_tid_tx_stats *total_tx,
struct cdp_tid_rx_stats *total_rx, uint8_t type)
{
uint8_t ring_id = 0, drop = 0;
struct cdp_tid_stats *tid_stats = &pdev->stats.tid_stats;
struct cdp_tid_tx_stats *per_ring_tx = NULL;
struct cdp_tid_rx_stats *per_ring_rx = NULL;
if (wlan_cfg_get_dp_soc_nss_cfg(pdev->soc->wlan_cfg_ctx)) {
qdf_mem_copy(total_tx, &tid_stats->tid_tx_stats[0][tid],
sizeof(struct cdp_tid_tx_stats));
qdf_mem_copy(total_rx, &tid_stats->tid_rx_stats[0][tid],
sizeof(struct cdp_tid_rx_stats));
return;
} else {
qdf_mem_zero(total_tx, sizeof(struct cdp_tid_tx_stats));
qdf_mem_zero(total_rx, sizeof(struct cdp_tid_rx_stats));
}
switch (type) {
case TID_COUNTER_STATS:
{
for (ring_id = 0; ring_id < CDP_MAX_TX_COMP_RINGS; ring_id++) {
per_ring_tx = &tid_stats->tid_tx_stats[ring_id][tid];
total_tx->success_cnt += per_ring_tx->success_cnt;
total_tx->comp_fail_cnt += per_ring_tx->comp_fail_cnt;
for (drop = 0; drop < TX_MAX_DROP; drop++)
total_tx->swdrop_cnt[drop] +=
per_ring_tx->swdrop_cnt[drop];
}
for (ring_id = 0; ring_id < CDP_MAX_RX_RINGS; ring_id++) {
per_ring_rx = &tid_stats->tid_rx_stats[ring_id][tid];
total_rx->delivered_to_stack +=
per_ring_rx->delivered_to_stack;
total_rx->intrabss_cnt += per_ring_rx->intrabss_cnt;
total_rx->msdu_cnt += per_ring_rx->msdu_cnt;
total_rx->mcast_msdu_cnt += per_ring_rx->mcast_msdu_cnt;
total_rx->bcast_msdu_cnt += per_ring_rx->bcast_msdu_cnt;
for (drop = 0; drop < RX_MAX_DROP; drop++)
total_rx->fail_cnt[drop] +=
per_ring_rx->fail_cnt[drop];
}
break;
}
case TID_DELAY_STATS:
{
for (ring_id = 0; ring_id < CDP_MAX_TX_COMP_RINGS; ring_id++) {
per_ring_tx = &tid_stats->tid_tx_stats[ring_id][tid];
dp_accumulate_delay_stats(&total_tx->swq_delay,
&per_ring_tx->swq_delay);
dp_accumulate_delay_stats(&total_tx->hwtx_delay,
&per_ring_tx->hwtx_delay);
dp_accumulate_delay_stats(&total_tx->intfrm_delay,
&per_ring_tx->intfrm_delay);
}
for (ring_id = 0; ring_id < CDP_MAX_RX_RINGS; ring_id++) {
per_ring_rx = &tid_stats->tid_rx_stats[ring_id][tid];
dp_accumulate_delay_stats(&total_rx->intfrm_delay,
&per_ring_rx->intfrm_delay);
dp_accumulate_delay_stats(&total_rx->to_stack_delay,
&per_ring_rx->to_stack_delay);
}
break;
}
default:
qdf_err("Invalid stats type");
break;
}
}
void dp_pdev_print_tid_stats(struct dp_pdev *pdev)
{
struct cdp_tid_stats *tid_stats;
struct cdp_tid_tx_stats *txstats;
struct cdp_tid_rx_stats *rxstats;
struct dp_soc *soc = pdev->soc;
struct cdp_tid_tx_stats total_tx;
struct cdp_tid_rx_stats total_rx;
uint8_t tid;
if (!soc)
return;
tid = 0;
rxstats = NULL;
txstats = NULL;
tid_stats = &pdev->stats.tid_stats;
DP_PRINT_STATS("Packets received in hardstart: %llu ",
tid_stats->ingress_stack);
pdev->stats.tid_stats.ingress_stack);
DP_PRINT_STATS("Packets dropped in osif layer: %llu ",
tid_stats->osif_drop);
pdev->stats.tid_stats.osif_drop);
DP_PRINT_STATS("Per TID Video Stats:\n");
for (tid = 0; tid < CDP_MAX_DATA_TIDS; tid++) {
txstats = &tid_stats->tid_tx_stats[tid];
rxstats = &tid_stats->tid_rx_stats[tid];
dp_accumulate_tid_stats(pdev, tid, &total_tx, &total_rx,
TID_COUNTER_STATS);
DP_PRINT_STATS("----TID: %d----", tid);
DP_PRINT_STATS("Tx Success Count: %llu", txstats->success_cnt);
DP_PRINT_STATS("Tx Complete Count: %llu",
txstats->complete_cnt);
DP_PRINT_STATS("Tx Success Count: %llu", total_tx.success_cnt);
DP_PRINT_STATS("Tx Firmware Drop Count: %llu",
txstats->comp_fail_cnt);
total_tx.comp_fail_cnt);
DP_PRINT_STATS("Tx Hardware Drop Count: %llu",
txstats->swdrop_cnt[TX_HW_ENQUEUE]);
total_tx.swdrop_cnt[TX_HW_ENQUEUE]);
DP_PRINT_STATS("Tx Software Drop Count: %llu",
txstats->swdrop_cnt[TX_SW_ENQUEUE]);
total_tx.swdrop_cnt[TX_SW_ENQUEUE]);
DP_PRINT_STATS("Tx Descriptor Error Count: %llu",
txstats->swdrop_cnt[TX_DESC_ERR]);
total_tx.swdrop_cnt[TX_DESC_ERR]);
DP_PRINT_STATS("Tx HAL Ring Error Count: %llu",
txstats->swdrop_cnt[TX_HAL_RING_ACCESS_ERR]);
total_tx.swdrop_cnt[TX_HAL_RING_ACCESS_ERR]);
DP_PRINT_STATS("Tx Dma Map Error Count: %llu",
txstats->swdrop_cnt[TX_DMA_MAP_ERR]);
total_tx.swdrop_cnt[TX_DMA_MAP_ERR]);
DP_PRINT_STATS("Rx Delievered Count: %llu",
rxstats->delivered_to_stack);
total_rx.delivered_to_stack);
DP_PRINT_STATS("Rx Software Enqueue Drop Count: %llu",
rxstats->fail_cnt[ENQUEUE_DROP]);
total_rx.fail_cnt[ENQUEUE_DROP]);
DP_PRINT_STATS("Rx Intrabss Drop Count: %llu",
rxstats->fail_cnt[INTRABSS_DROP]);
total_rx.fail_cnt[INTRABSS_DROP]);
DP_PRINT_STATS("Rx Msdu Done Failure Count: %llu",
rxstats->fail_cnt[MSDU_DONE_FAILURE]);
total_rx.fail_cnt[MSDU_DONE_FAILURE]);
DP_PRINT_STATS("Rx Invalid Peer Count: %llu",
rxstats->fail_cnt[INVALID_PEER_VDEV]);
total_rx.fail_cnt[INVALID_PEER_VDEV]);
DP_PRINT_STATS("Rx Policy Check Drop Count: %llu",
rxstats->fail_cnt[POLICY_CHECK_DROP]);
total_rx.fail_cnt[POLICY_CHECK_DROP]);
DP_PRINT_STATS("Rx Mec Drop Count: %llu",
rxstats->fail_cnt[MEC_DROP]);
total_rx.fail_cnt[MEC_DROP]);
DP_PRINT_STATS("Rx Nawds Mcast Drop Count: %llu",
rxstats->fail_cnt[NAWDS_MCAST_DROP]);
total_rx.fail_cnt[NAWDS_MCAST_DROP]);
DP_PRINT_STATS("Rx Mesh Filter Drop Count: %llu",
rxstats->fail_cnt[MESH_FILTER_DROP]);
total_rx.fail_cnt[MESH_FILTER_DROP]);
DP_PRINT_STATS("Rx Intra Bss Deliver Count: %llu",
rxstats->intrabss_cnt);
DP_PRINT_STATS("Rx MSDU Count: %llu", rxstats->msdu_cnt);
total_rx.intrabss_cnt);
DP_PRINT_STATS("Rx MSDU Count: %llu", total_rx.msdu_cnt);
DP_PRINT_STATS("Rx Multicast MSDU Count: %llu",
rxstats->mcast_msdu_cnt);
total_rx.mcast_msdu_cnt);
DP_PRINT_STATS("Rx Broadcast MSDU Count: %llu\n",
rxstats->bcast_msdu_cnt);
total_rx.bcast_msdu_cnt);
}
}
void dp_pdev_print_delay_stats(struct dp_pdev *pdev)
{
struct dp_soc *soc = pdev->soc;
struct cdp_tid_tx_stats *txstats = NULL;
struct cdp_tid_rx_stats *rxstats;
struct cdp_tid_tx_stats total_tx;
struct cdp_tid_rx_stats total_rx;
struct cdp_tid_stats *tid_stats;
uint8_t tid, index;
uint64_t count = 0;
@@ -4111,17 +4209,17 @@ void dp_pdev_print_delay_stats(struct dp_pdev *pdev)
tid = 0;
index = 0;
rxstats = NULL;
tid_stats = &pdev->stats.tid_stats;
DP_PRINT_STATS("Per TID Delay Non-Zero Stats:\n");
for (tid = 0; tid < CDP_MAX_DATA_TIDS; tid++) {
dp_accumulate_tid_stats(pdev, tid, &total_tx, &total_rx,
TID_DELAY_STATS);
DP_PRINT_STATS("----TID: %d----", tid);
txstats = &pdev->stats.tid_stats.tid_tx_stats[tid];
rxstats = &pdev->stats.tid_stats.tid_rx_stats[tid];
DP_PRINT_STATS("Software Enqueue Delay:");
for (index = 0; index < CDP_DELAY_BUCKET_MAX; index++) {
count = txstats->swq_delay.delay_bucket[index];
count = total_tx.swq_delay.delay_bucket[index];
if (count) {
DP_PRINT_STATS("%s: Packets = %llu",
dp_vow_str_sw_enq_delay(index),
@@ -4129,52 +4227,52 @@ void dp_pdev_print_delay_stats(struct dp_pdev *pdev)
}
}
DP_PRINT_STATS("Min = %u", txstats->swq_delay.min_delay);
DP_PRINT_STATS("Max = %u", txstats->swq_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", txstats->swq_delay.avg_delay);
DP_PRINT_STATS("Min = %u", total_tx.swq_delay.min_delay);
DP_PRINT_STATS("Max = %u", total_tx.swq_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", total_tx.swq_delay.avg_delay);
DP_PRINT_STATS("Hardware Transmission Delay:");
for (index = 0; index < CDP_DELAY_BUCKET_MAX; index++) {
count = txstats->hwtx_delay.delay_bucket[index];
count = total_tx.hwtx_delay.delay_bucket[index];
if (count) {
DP_PRINT_STATS("%s: Packets = %llu",
dp_vow_str_fw_to_hw_delay(index),
count);
}
}
DP_PRINT_STATS("Min = %u", txstats->hwtx_delay.min_delay);
DP_PRINT_STATS("Max = %u", txstats->hwtx_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", txstats->hwtx_delay.avg_delay);
DP_PRINT_STATS("Min = %u", total_tx.hwtx_delay.min_delay);
DP_PRINT_STATS("Max = %u", total_tx.hwtx_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", total_tx.hwtx_delay.avg_delay);
DP_PRINT_STATS("Tx Interframe Delay:");
for (index = 0; index < CDP_DELAY_BUCKET_MAX; index++) {
count = txstats->intfrm_delay.delay_bucket[index];
count = total_tx.intfrm_delay.delay_bucket[index];
if (count) {
DP_PRINT_STATS("%s: Packets = %llu",
dp_vow_str_intfrm_delay(index),
count);
}
}
DP_PRINT_STATS("Min = %u", txstats->intfrm_delay.min_delay);
DP_PRINT_STATS("Max = %u", txstats->intfrm_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", txstats->intfrm_delay.avg_delay);
DP_PRINT_STATS("Min = %u", total_tx.intfrm_delay.min_delay);
DP_PRINT_STATS("Max = %u", total_tx.intfrm_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", total_tx.intfrm_delay.avg_delay);
DP_PRINT_STATS("Rx Interframe Delay:");
for (index = 0; index < CDP_DELAY_BUCKET_MAX; index++) {
count = rxstats->intfrm_delay.delay_bucket[index];
count = total_rx.intfrm_delay.delay_bucket[index];
if (count) {
DP_PRINT_STATS("%s: Packets = %llu",
dp_vow_str_intfrm_delay(index),
count);
}
}
DP_PRINT_STATS("Min = %u", rxstats->intfrm_delay.min_delay);
DP_PRINT_STATS("Max = %u", rxstats->intfrm_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", rxstats->intfrm_delay.avg_delay);
DP_PRINT_STATS("Min = %u", total_rx.intfrm_delay.min_delay);
DP_PRINT_STATS("Max = %u", total_rx.intfrm_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", total_rx.intfrm_delay.avg_delay);
DP_PRINT_STATS("Rx Reap to Stack Delay:");
for (index = 0; index < CDP_DELAY_BUCKET_MAX; index++) {
count = rxstats->to_stack_delay.delay_bucket[index];
count = total_rx.to_stack_delay.delay_bucket[index];
if (count) {
DP_PRINT_STATS("%s: Packets = %llu",
dp_vow_str_intfrm_delay(index),
@@ -4182,9 +4280,9 @@ void dp_pdev_print_delay_stats(struct dp_pdev *pdev)
}
}
DP_PRINT_STATS("Min = %u", rxstats->to_stack_delay.min_delay);
DP_PRINT_STATS("Max = %u", rxstats->to_stack_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", rxstats->to_stack_delay.avg_delay);
DP_PRINT_STATS("Min = %u", total_rx.to_stack_delay.min_delay);
DP_PRINT_STATS("Max = %u", total_rx.to_stack_delay.max_delay);
DP_PRINT_STATS("Avg = %u\n", total_rx.to_stack_delay.avg_delay);
}
}
#endif

View File

@@ -1455,7 +1455,8 @@ static qdf_nbuf_t dp_tx_send_msdu_single(struct dp_vdev *vdev, qdf_nbuf_t nbuf,
dp_err_rl("Tx_desc prepare Fail vdev %pK queue %d",
vdev, tx_q->desc_pool_id);
dp_tx_get_tid(vdev, nbuf, msdu_info);
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[msdu_info->tid];
tid_stats = &pdev->stats.tid_stats.
tid_tx_stats[tx_q->ring_id][msdu_info->tid];
tid_stats->swdrop_cnt[TX_DESC_ERR]++;
return nbuf;
}
@@ -1475,7 +1476,8 @@ static qdf_nbuf_t dp_tx_send_msdu_single(struct dp_vdev *vdev, qdf_nbuf_t nbuf,
"%s %d : HAL RING Access Failed -- %pK",
__func__, __LINE__, hal_srng);
dp_tx_get_tid(vdev, nbuf, msdu_info);
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[msdu_info->tid];
tid_stats = &pdev->stats.tid_stats.
tid_tx_stats[tx_q->ring_id][tid];
tid_stats->swdrop_cnt[TX_HAL_RING_ACCESS_ERR]++;
DP_STATS_INC(vdev, tx_i.dropped.ring_full, 1);
dp_tx_desc_release(tx_desc, tx_q->desc_pool_id);
@@ -1508,7 +1510,8 @@ static qdf_nbuf_t dp_tx_send_msdu_single(struct dp_vdev *vdev, qdf_nbuf_t nbuf,
"%s Tx_hw_enqueue Fail tx_desc %pK queue %d",
__func__, tx_desc, tx_q->ring_id);
dp_tx_get_tid(vdev, nbuf, msdu_info);
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[msdu_info->tid];
tid_stats = &pdev->stats.tid_stats.
tid_tx_stats[tx_q->ring_id][tid];
tid_stats->swdrop_cnt[TX_HW_ENQUEUE]++;
dp_tx_desc_release(tx_desc, tx_q->desc_pool_id);
qdf_nbuf_unmap(vdev->osdev, nbuf, QDF_DMA_TO_DEVICE);
@@ -1564,7 +1567,8 @@ qdf_nbuf_t dp_tx_send_msdu_multiple(struct dp_vdev *vdev, qdf_nbuf_t nbuf,
"%s %d : HAL RING Access Failed -- %pK",
__func__, __LINE__, hal_srng);
dp_tx_get_tid(vdev, nbuf, msdu_info);
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[msdu_info->tid];
tid_stats = &pdev->stats.tid_stats.
tid_tx_stats[tx_q->ring_id][msdu_info->tid];
tid_stats->swdrop_cnt[TX_HAL_RING_ACCESS_ERR]++;
DP_STATS_INC(vdev, tx_i.dropped.ring_full, 1);
return nbuf;
@@ -1631,7 +1635,7 @@ qdf_nbuf_t dp_tx_send_msdu_multiple(struct dp_vdev *vdev, qdf_nbuf_t nbuf,
dp_tx_get_tid(vdev, nbuf, msdu_info);
tid_stats = &pdev->stats.tid_stats.
tid_tx_stats[msdu_info->tid];
tid_tx_stats[tx_q->ring_id][msdu_info->tid];
tid_stats->swdrop_cnt[TX_HW_ENQUEUE]++;
if (tx_desc->flags & DP_TX_DESC_FLAG_ME)
dp_tx_me_free_buf(pdev, tx_desc->me_buffer);
@@ -2665,10 +2669,12 @@ void dp_tx_comp_fill_tx_completion_stats(struct dp_tx_desc_s *tx_desc,
* @vdev: pdev handle
* @tx_desc: tx descriptor
* @tid: tid value
* @ring_id: TCL or WBM ring number for transmit path
* Return: none
*/
static void dp_tx_compute_delay(struct dp_vdev *vdev,
struct dp_tx_desc_s *tx_desc, uint8_t tid)
struct dp_tx_desc_s *tx_desc,
uint8_t tid, uint8_t ring_id)
{
int64_t current_timestamp, timestamp_ingress, timestamp_hw_enqueue;
uint32_t sw_enqueue_delay, fwhw_transmit_delay, interframe_delay;
@@ -2689,12 +2695,12 @@ static void dp_tx_compute_delay(struct dp_vdev *vdev,
* Delay in software enqueue
*/
dp_update_delay_stats(vdev->pdev, sw_enqueue_delay, tid,
CDP_DELAY_STATS_SW_ENQ);
CDP_DELAY_STATS_SW_ENQ, ring_id);
/*
* Delay between packet enqueued to HW and Tx completion
*/
dp_update_delay_stats(vdev->pdev, fwhw_transmit_delay, tid,
CDP_DELAY_STATS_FW_HW_TRANSMIT);
CDP_DELAY_STATS_FW_HW_TRANSMIT, ring_id);
/*
* Update interframe delay stats calculated at hardstart receive point.
@@ -2704,22 +2710,25 @@ static void dp_tx_compute_delay(struct dp_vdev *vdev,
* of !vdev->prev_tx_enq_tstamp.
*/
dp_update_delay_stats(vdev->pdev, interframe_delay, tid,
CDP_DELAY_STATS_TX_INTERFRAME);
CDP_DELAY_STATS_TX_INTERFRAME, ring_id);
vdev->prev_tx_enq_tstamp = timestamp_ingress;
}
/**
* dp_tx_update_peer_stats() - Update peer stats from Tx completion indications
* per wbm ring
*
* @tx_desc: software descriptor head pointer
* @ts: Tx completion status
* @peer: peer handle
* @ring_id: ring number
*
* Return: None
*/
static inline void
dp_tx_update_peer_stats(struct dp_tx_desc_s *tx_desc,
struct hal_tx_completion_status *ts,
struct dp_peer *peer)
struct dp_peer *peer, uint8_t ring_id)
{
struct dp_pdev *pdev = peer->vdev->pdev;
struct dp_soc *soc = NULL;
@@ -2734,7 +2743,7 @@ dp_tx_update_peer_stats(struct dp_tx_desc_s *tx_desc,
if (qdf_unlikely(tid >= CDP_MAX_DATA_TIDS))
tid = CDP_MAX_DATA_TIDS - 1;
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[tid];
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[ring_id][tid];
soc = pdev->soc;
mcs = ts->mcs;
@@ -2749,8 +2758,7 @@ dp_tx_update_peer_stats(struct dp_tx_desc_s *tx_desc,
DP_STATS_INC_PKT(peer, tx.comp_pkt, 1, length);
if (qdf_unlikely(pdev->delay_stats_flag))
dp_tx_compute_delay(peer->vdev, tx_desc, tid);
tid_stats->complete_cnt++;
dp_tx_compute_delay(peer->vdev, tx_desc, tid, ring_id);
DP_STATS_INCC(peer, tx.dropped.age_out, 1,
(ts->status == HAL_TX_TQM_RR_REM_CMD_AGED));
@@ -3023,13 +3031,14 @@ dp_tx_comp_process_desc(struct dp_soc *soc,
* @tx_desc: software descriptor head pointer
* @ts: Tx completion status
* @peer: peer handle
* @ring_id: ring number
*
* Return: none
*/
static inline
void dp_tx_comp_process_tx_status(struct dp_tx_desc_s *tx_desc,
struct hal_tx_completion_status *ts,
struct dp_peer *peer)
struct dp_peer *peer, uint8_t ring_id)
{
uint32_t length;
qdf_ether_header_t *eh;
@@ -3119,7 +3128,7 @@ void dp_tx_comp_process_tx_status(struct dp_tx_desc_s *tx_desc,
}
}
dp_tx_update_peer_stats(tx_desc, ts, peer);
dp_tx_update_peer_stats(tx_desc, ts, peer, ring_id);
#ifdef QCA_SUPPORT_RDK_STATS
if (soc->wlanstats_enabled)
@@ -3135,6 +3144,7 @@ out:
* dp_tx_comp_process_desc_list() - Tx complete software descriptor handler
* @soc: core txrx main context
* @comp_head: software descriptor head pointer
* @ring_id: ring number
*
* This function will process batch of descriptors reaped by dp_tx_comp_handler
* and release the software descriptors after processing is complete
@@ -3143,7 +3153,7 @@ out:
*/
static void
dp_tx_comp_process_desc_list(struct dp_soc *soc,
struct dp_tx_desc_s *comp_head)
struct dp_tx_desc_s *comp_head, uint8_t ring_id)
{
struct dp_tx_desc_s *desc;
struct dp_tx_desc_s *next;
@@ -3156,7 +3166,7 @@ dp_tx_comp_process_desc_list(struct dp_soc *soc,
while (desc) {
hal_tx_comp_get_status(&desc->comp, &ts, soc->hal_soc);
peer = dp_peer_find_by_id(soc, ts.peer_id);
dp_tx_comp_process_tx_status(desc, &ts, peer);
dp_tx_comp_process_tx_status(desc, &ts, peer, ring_id);
netbuf = desc->nbuf;
/* check tx complete notification */
@@ -3180,13 +3190,15 @@ dp_tx_comp_process_desc_list(struct dp_soc *soc,
* dp_tx_process_htt_completion() - Tx HTT Completion Indication Handler
* @tx_desc: software descriptor head pointer
* @status : Tx completion status from HTT descriptor
* @ring_id: ring number
*
* This function will process HTT Tx indication messages from Target
*
* Return: none
*/
static
void dp_tx_process_htt_completion(struct dp_tx_desc_s *tx_desc, uint8_t *status)
void dp_tx_process_htt_completion(struct dp_tx_desc_s *tx_desc, uint8_t *status,
uint8_t ring_id)
{
uint8_t tx_status;
struct dp_pdev *pdev;
@@ -3238,11 +3250,10 @@ void dp_tx_process_htt_completion(struct dp_tx_desc_s *tx_desc, uint8_t *status)
if (qdf_unlikely(tid >= CDP_MAX_DATA_TIDS))
tid = CDP_MAX_DATA_TIDS - 1;
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[tid];
tid_stats = &pdev->stats.tid_stats.tid_tx_stats[ring_id][tid];
if (qdf_unlikely(pdev->delay_stats_flag))
dp_tx_compute_delay(vdev, tx_desc, tid);
tid_stats->complete_cnt++;
dp_tx_compute_delay(vdev, tx_desc, tid, ring_id);
if (qdf_unlikely(tx_status != HTT_TX_FW2WBM_TX_STATUS_OK)) {
ts.status = HAL_TX_TQM_RR_REM_CMD_REM;
tid_stats->comp_fail_cnt++;
@@ -3255,7 +3266,7 @@ void dp_tx_process_htt_completion(struct dp_tx_desc_s *tx_desc, uint8_t *status)
if (qdf_likely(peer))
dp_peer_unref_del_find_by_id(peer);
dp_tx_comp_process_tx_status(tx_desc, &ts, peer);
dp_tx_comp_process_tx_status(tx_desc, &ts, peer, ring_id);
dp_tx_comp_process_desc(soc, tx_desc, &ts, peer);
dp_tx_desc_release(tx_desc, tx_desc->pool_id);
@@ -3318,7 +3329,7 @@ static inline bool dp_tx_comp_enable_eol_data_check(struct dp_soc *soc)
#endif
uint32_t dp_tx_comp_handler(struct dp_intr *int_ctx, struct dp_soc *soc,
void *hal_srng, uint32_t quota)
void *hal_srng, uint8_t ring_id, uint32_t quota)
{
void *tx_comp_hal_desc;
uint8_t buffer_src;
@@ -3400,7 +3411,7 @@ more_data:
hal_tx_comp_get_htt_desc(tx_comp_hal_desc,
htt_tx_status);
dp_tx_process_htt_completion(tx_desc,
htt_tx_status);
htt_tx_status, ring_id);
} else {
/* Pool id is not matching. Error */
if (tx_desc->pool_id != pool_id) {
@@ -3460,7 +3471,7 @@ more_data:
/* Process the reaped descriptors */
if (head_desc)
dp_tx_comp_process_desc_list(soc, head_desc);
dp_tx_comp_process_desc_list(soc, head_desc, ring_id);
if (dp_tx_comp_enable_eol_data_check(soc)) {
if (!force_break &&

View File

@@ -185,6 +185,7 @@ qdf_nbuf_t dp_tx_non_std(struct cdp_vdev *vdev_handle,
* dp_tx_comp_handler() - Tx completion handler
* @int_ctx: pointer to DP interrupt context
* @soc: core txrx main context
* @hal_srng: Opaque HAL SRNG pointer
* @ring_id: completion ring id
* @quota: No. of packets/descriptors that can be serviced in one loop
*
@@ -195,7 +196,7 @@ qdf_nbuf_t dp_tx_non_std(struct cdp_vdev *vdev_handle,
* Return: Number of TX completions processed
*/
uint32_t dp_tx_comp_handler(struct dp_intr *int_ctx, struct dp_soc *soc,
void *hal_srng, uint32_t quota);
void *hal_srng, uint8_t ring_id, uint32_t quota);
QDF_STATUS
dp_tx_prepare_send_me(struct dp_vdev *vdev, qdf_nbuf_t nbuf);