wifi: ath12k: extend device DP stats for TX and RX observability
cover wifi: ath12k: extend device DP stats for TX and RX observability
From: Pardeep Kaur <[email protected]>
Extend ath12k device DP stats for TX and RX path observability.
Patch 1 fixes out-of-bounds array access on TX stats arrays fw_tx_status[], tx_wbm_rel_source[] and tqm_rel_reason[] by adding likely()/WARN_ON_ONCE() bounds checks before each increment.
Patch 2 renames the local variable wbm_status to htt_status in ath12k_dp_tx_process_htt_tx_complete() to accurately reflect the field origin, and simplifies ts.acked to use true directly.
Patch 3 adds a per-TCL-ring TX buffer allocation failure counter txbuf_na[] to track how often ath12k_dp_tx_assign_buffer() fails, and exposes it in device_dp_stats debugfs.
Patch 4 adds a write handler to device_dp_stats debugfs so that writing 'reset' clears all device DP stats counters without reloading the driver.
Patch 5 introduces a new dp_stats.h header with the ath12k_wbm_err_drop_reason enum, adds per-drop-reason counters in the WBM RX error path, and exposes them in device_dp_stats debugfs.
Patch 6 adds a per-REO-ring sent_to_stack[] counter to track successful MSDU deliveries to the network stack, and exposes it in device_dp_stats debugfs.
Patch 7 fixes the REO Rx Received debugfs output to use 0-based ring indexing and lowercase "ring%d:" consistent with other ring counters.
Patch 8 adds a dedicated sw_desc_fallback counter to track how often the WBM RX error path falls back to SW descriptor retrieval when HW CC is not done, distinct from actual packet drops.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
---
Aniruddha Mishra (2):
wifi: ath12k: add WBM RX error drop statistics
wifi: ath12k: add WBM SW desc fallback counter
Hariharan Ramanathan (4):
wifi: ath12k: rename wbm_status to htt_status in HTT TX completion
wifi: ath12k: add TCL ring TX buffer allocation failure counter
wifi: ath12k: add device DP stats reset support via debugfs
wifi: ath12k: track per-ring RX sent-to-stack count
Pardeep Kaur (2):
wifi: ath12k: fix out-of-bounds access on TX stats arrays
wifi: ath12k: fix 1-based ring index in REO Rx Received debugfs output
drivers/net/wireless/ath/ath12k/debugfs.c | 74 ++++++++++++++++++- drivers/net/wireless/ath/ath12k/dp.h | 14 +++- drivers/net/wireless/ath/ath12k/dp_stats.h | 25 +++++++ drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c | 52 ++++++++++--- drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c | 29 +++++--- .../net/wireless/ath/ath12k/wifi7/hal_rx.c | 5 +- 6 files changed, 173 insertions(+), 26 deletions(-) create mode 100644 drivers/net/wireless/ath/ath12k/dp_stats.h
base-commit: 32f39e331f889a0424f7a0a82893f628e2705727 --
2.34.1
PATCH 1/8 wifi: ath12k: fix out-of-bounds access on TX stats arrays
From: Pardeep Kaur <[email protected]>
The fw_tx_status[], tx_wbm_rel_source[], and tqm_rel_reason[] arrays are indexed directly by values read from hardware without bounds checks. Values at or beyond MAX_FW_TX_STATUS, HAL_WBM_REL_SRC_MODULE_MAX, or MAX_TQM_RELEASE_REASON respectively would write past the end of the arrays causing memory corruption.
Add bounds checks using likely() and WARN_ON_ONCE() before incrementing each counter, consistent with the existing pattern used elsewhere in the ath12k codebase.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-)
drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c +12 -3
@@ -520,7 +520,10 @@ ath12k_dp_tx_process_htt_tx_complete(struct ath12k_dp *dp, void *desc, wbm_status =le32_get_bits (status_desc->info0,HTT_TX_WBM_COMP_INFO0_STATUS ); - dp->device_stats.fw_tx_status[wbm_status]++; +if (likely (wbm_status <MAX_FW_TX_STATUS )) + dp->device_stats.fw_tx_status[wbm_status]++; +else +WARN_ON_ONCE (1 );switch (wbm_status) {case HAL_WBM_REL_HTT_TX_COMP_STATUS_OK :
@@ -922,11 +925,17 @@ void ath12k_wifi7_dp_tx_completion_handler(struct ath12k_dp *dp, int ring_id)/* Find the HAL_WBM_RELEASE_INFO0_REL_SRC_MODULE value */ buf_rel_source =le32_get_bits (tx_status->info0,HAL_WBM_RELEASE_INFO0_REL_SRC_MODULE ); - dp->device_stats.tx_wbm_rel_source[buf_rel_source]++; +if (likely (buf_rel_source <HAL_WBM_REL_SRC_MODULE_MAX )) + dp->device_stats.tx_wbm_rel_source[buf_rel_source]++; +else +WARN_ON_ONCE (1 ); rel_status =le32_get_bits (tx_status->info0,HAL_WBM_COMPL_TX_INFO0_TQM_RELEASE_REASON ); - dp->device_stats.tqm_rel_reason[rel_status]++; +if (likely (rel_status <MAX_TQM_RELEASE_REASON )) + dp->device_stats.tqm_rel_reason[rel_status]++; +else +WARN_ON_ONCE (1 );/* Release descriptor as soon as extracting necessary info * to reduce contention
base-commit: 32f39e331f889a0424f7a0a82893f628e2705727 --
2.34.1
PATCH 2/8 wifi: ath12k: rename wbm_status to htt_status in HTT TX completion
In reply to Pardeep KaurFrom: Hariharan Ramanathan <[email protected]>
The variable wbm_status in ath12k_dp_tx_process_htt_tx_complete() holds a value from the HTT TX WBM completion status field, not a generic WBM status. Rename it to htt_status to accurately reflect the field origin and improve code clarity.
Also fix ts.acked assignment inside the HAL_WBM_REL_HTT_TX_COMP_STATUS_OK case to use true directly since the condition is always true within that case branch.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c +7 -7
@@ -513,21 +513,21 @@ ath12k_dp_tx_process_htt_tx_complete(struct ath12k_dp *dp, void *desc, {struct htt_tx_wbm_completion *status_desc;struct ath12k_dp_htt_wbm_tx_status ts = {}; -enum hal_wbm_htt_tx_comp_status wbm_status; +enum hal_wbm_htt_tx_comp_status htt_status;u16 peer_id; status_desc = desc; - wbm_status =le32_get_bits (status_desc->info0, + htt_status =le32_get_bits (status_desc->info0,HTT_TX_WBM_COMP_INFO0_STATUS ); -if (likely (wbm_status <MAX_FW_TX_STATUS )) - dp->device_stats.fw_tx_status[wbm_status]++; +if (likely (htt_status <MAX_FW_TX_STATUS )) + dp->device_stats.fw_tx_status[htt_status]++;else WARN_ON_ONCE (1 ); -switch (wbm_status) { +switch (htt_status) {case HAL_WBM_REL_HTT_TX_COMP_STATUS_OK : - ts.acked = (wbm_status ==HAL_WBM_REL_HTT_TX_COMP_STATUS_OK ); + ts.acked = true; ts.ack_rssi =le32_get_bits (status_desc->info2,HTT_TX_WBM_COMP_INFO2_ACK_RSSI );
@@ -549,7 +549,7 @@ ath12k_dp_tx_process_htt_tx_complete(struct ath12k_dp *dp, void *desc,*/ break ;default : -ath12k_warn (dp->ab,"Unknown htt wbm tx status %d\n" , wbm_status); +ath12k_warn (dp->ab,"Unknown htt tx status %d\n" , htt_status);break ; } }
Signature · 1 line
2.34.1
PATCH 3/8 wifi: ath12k: add TCL ring TX buffer allocation failure counter
In reply to Pardeep KaurFrom: Hariharan Ramanathan <[email protected]>
When ath12k_dp_tx_assign_buffer() fails, the TX path returns -ENOMEM silently with no per-ring visibility into how often this occurs. Without a counter, buffer pool exhaustion on a specific TCL ring is invisible during debugging.
Add txbuf_na[] to ath12k_device_dp_tx_err_stats to track per-ring TX buffer allocation failures and increment it on assign failure in the TX path. Expose the per-ring counts in debugfs under a new 'TCL Ring Buffer Alloc Failures' section.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/debugfs.c | 5 +++++ drivers/net/wireless/ath/ath12k/dp.h | 2 ++ drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c | 4 +++- 3 files changed, 10 insertions(+), 1 deletion(-)
drivers/net/wireless/ath/ath12k/debugfs.c +5 -0
@@ -1117,6 +1117,11 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file, len +=scnprintf (buf + len, size - len,"ring%d: %u\n" , i, device_stats->tx_err.desc_na[i]); + len +=scnprintf (buf + len, size - len,"\nTCL Ring Buffer Alloc Failures:\n" ); +for (i =0 ; i <DP_TCL_NUM_RING_MAX ; i++) + len +=scnprintf (buf + len, size - len,"ring%d: %u\n" , + i, device_stats->tx_err.txbuf_na[i]); + len +=scnprintf (buf + len, size - len,"\nMisc Transmit Failures: %d\n" ,atomic_read (&device_stats->tx_err.misc_fail));
drivers/net/wireless/ath/ath12k/dp.h +2 -0
@@ -428,6 +428,8 @@ struct ath12k_dp_arch_ops {struct ath12k_device_dp_tx_err_stats {/* TCL Ring Descriptor unavailable */ u32 desc_na[DP_TCL_NUM_RING_MAX ]; +/* TCL Ring Buffers unavailable */ +u32 txbuf_na[DP_TCL_NUM_RING_MAX ];/* Other failures during dp_tx due to mem allocation failure * idr unavailable etc. */
drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c +3 -1
@@ -123,8 +123,10 @@ int ath12k_wifi7_dp_tx(struct ath12k_pdev_dp *dp_pdev, struct ath12k_link_vif *a tx_ring = &dp->tx_ring[ti.ring_id]; tx_desc =ath12k_dp_tx_assign_buffer (dp, pool_id); -if (!tx_desc) +if (!tx_desc) { + dp->device_stats.tx_err.txbuf_na[ti.ring_id]++;return -ENOMEM ; + } dp_link_vif =ath12k_dp_vif_to_dp_link_vif (&ahvif->dp_vif, arvif->link_id);
Signature · 1 line
2.34.1
Re: [PATCH ath-next 3/8] wifi: ath12k: add TCL ring TX buffer allocation failure counter
Quoted 11 lines · Pardeep Kaur
From: Hariharan Ramanathan <[email protected]>
When
ath12k_dp_tx_assign_buffer()fails, the TX path returns-ENOMEMsilently with no per-ring visibility into how often this occurs. Without a counter, buffer pool exhaustion on a specific TCL ring is invisible during debugging.Add txbuf_na[] to
ath12k_device_dp_tx_err_statsto track per-ring TX buffer allocation failures and increment it on assign failure in the TX path. Expose the per-ring counts in debugfs under a new 'TCL Ring Buffer Alloc Failures' section.
Is this counting the right thing?
One of my review agents has the observation:
The txbuf_na counter is indexed by ring_id but the exhaustion it measures is from tx_desc[pool_id]
So I drilled down with another agent and got:
The commit message intent is to answer "why was this frame dropped", which is valid. The practical question is whether ring_id or pool_id is more useful for diagnosis:
- pool_id = skb_get_queue_mapping(skb) & 3 — a property of the traffic
class/queue
- ring_id = smp_processor_id() % max_tx_ring — a property of which CPU
processed it
If a pool is exhausted, pool_id tells you which traffic class is backpressured. ring_id tells you which CPU happened to observe it — less actionable. So the other finding is a genuine semantic improvement, not just pedantry. The fix (txbuf_na[pool_id]++) would make the stat correctly answer "which traffic class ran out of TX descriptors."
So does tracking by ring_id or pool_id make the most sense?
Quoted 57 lines
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
Signed-off-by Hariharan Ramanathan <[email protected]>Co-developed-by Pardeep Kaur <[email protected]>Signed-off-by Pardeep Kaur <[email protected]>--- drivers/net/wireless/ath/ath12k/debugfs.c | 5 +++++ drivers/net/wireless/ath/ath12k/dp.h | 2 ++ drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c | 4 +++- 3 files changed, 10 insertions(+), 1 deletion(-)drivers/net/wireless/ath/ath12k/debugfs.c +5
-0@@ -1117,6 +1117,11 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file, len +=scnprintf (buf + len, size - len,"ring%d: %u\n" , i, device_stats->tx_err.desc_na[i]); + len +=scnprintf (buf + len, size - len,"\nTCL Ring Buffer Alloc Failures:\n" ); +for (i =0 ; i <DP_TCL_NUM_RING_MAX ; i++) + len +=scnprintf (buf + len, size - len,"ring%d: %u\n" , + i, device_stats->tx_err.txbuf_na[i]); + len +=scnprintf (buf + len, size - len,"\nMisc Transmit Failures: %d\n" ,atomic_read (&device_stats->tx_err.misc_fail));drivers/net/wireless/ath/ath12k/dp.h +2
-0@@ -428,6 +428,8 @@ struct ath12k_dp_arch_ops {struct ath12k_device_dp_tx_err_stats {/* TCL Ring Descriptor unavailable */ u32 desc_na[DP_TCL_NUM_RING_MAX ]; +/* TCL Ring Buffers unavailable */ +u32 txbuf_na[DP_TCL_NUM_RING_MAX ];/* Other failures during dp_tx due to mem allocation failure * idr unavailable etc. */ drivers/net/wireless/ath/ath12k/wifi7/dp_tx.c +3
-1@@ -123,8 +123,10 @@ int ath12k_wifi7_dp_tx(struct ath12k_pdev_dp *dp_pdev, struct ath12k_link_vif *a tx_ring = &dp->tx_ring[ti.ring_id]; tx_desc =ath12k_dp_tx_assign_buffer (dp, pool_id); -if (!tx_desc) +if (!tx_desc) { + dp->device_stats.tx_err.txbuf_na[ti.ring_id]++;return -ENOMEM ; + } dp_link_vif =ath12k_dp_vif_to_dp_link_vif (&ahvif->dp_vif, arvif->link_id);
PATCH 4/8 wifi: ath12k: add device DP stats reset support via debugfs
In reply to Pardeep KaurFrom: Hariharan Ramanathan <[email protected]>
There is no way to reset device DP stats counters without reloading the driver, making it difficult to isolate issues to a specific time window during debugging.
Add a write handler to the device_dp_stats debugfs file so that writing 'reset' clears all device DP stats counters. Change the file mode from 0400 to 0600 to allow write access. Use strcmp() with newline stripping to match the input, consistent with the existing pattern in debugfs.c. Return -EINVAL on unrecognised input.
No lock is taken around the memset since the counters are updated locklessly in the datapath; taking dp_lock would be misleading as it does not protect device_stats updates.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/debugfs.c | 35 ++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-)
drivers/net/wireless/ath/ath12k/debugfs.c +34 -1
@@ -1220,8 +1220,41 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file,return ret; } +static ssize_t +ath12k_debugfs_write_device_dp_stats (struct file *file, +const char __user *user_buf, +size_t count,loff_t *ppos) +{ +struct ath12k_base *ab = file->private_data; +struct ath12k_dp *dp =ath12k_ab_to_dp (ab); +struct ath12k_device_dp_stats *device_stats = &dp->device_stats; +char buf[20 ] = {}; +int ret; + +if (count >=sizeof (buf)) +return -EINVAL ; + + ret =copy_from_user (buf, user_buf, count); +if (ret) +return -EFAULT ; + +/* drop the possible '\n' from the end */ +if (count >0 && buf[count -1 ] =='\n' ) + buf[count -1 ] ='\0' ; + +if (!strcmp (buf,"reset" )) { +memset (device_stats,0 ,sizeof (*device_stats)); +return count; + } + +ath12k_warn (ab,"unsupported command: %s\n" , buf); + +return -EINVAL ; +} +static const struct file_operations fops_device_dp_stats = { .read = ath12k_debugfs_dump_device_dp_stats, + .write = ath12k_debugfs_write_device_dp_stats, .open = simple_open, .owner =THIS_MODULE , .llseek = default_llseek,
@@ -1232,7 +1265,7 @@ void ath12k_debugfs_pdev_create(struct ath12k_base *ab)debugfs_create_file ("simulate_fw_crash" ,0600 , ab->debugfs_soc, ab, &fops_simulate_fw_crash); -debugfs_create_file ("device_dp_stats" ,0400 , ab->debugfs_soc, ab, +debugfs_create_file ("device_dp_stats" ,0600 , ab->debugfs_soc, ab, &fops_device_dp_stats); }
Signature · 1 line
2.34.1
PATCH 5/8 wifi: ath12k: add WBM RX error drop statistics
In reply to Pardeep KaurFrom: Aniruddha Mishra <[email protected]>
Without fine-grained drop counters, diagnosing RX failures in the WBM error path requires intrusive debugging. Introduce per-drop-reason counters to track exactly where and why packets are dropped during WBM RX error handling.
Introduce a new dp_stats.h header and define ath12k_wbm_err_drop_reason enum there covering all drop points in the WBM RX error path: descriptor parse failures, SW descriptor retrieval errors, null partner DP, invalid HW link ID, null pdev/ar, CAC running, scatter-gather drops, and invalid NWifi header length. The new header is included by dp.h which is already included by all consumers.
Add ath12k_device_dp_rx_wbm_err_stats struct grouping rxdma_error[], reo_error[], and the new drop[] counters, replacing the flat arrays previously in ath12k_device_dp_stats.
Add bounds checks before indexing rxdma_error[] and reo_error[] with hardware-supplied err_code values to prevent out-of-bounds writes, using the same likely/WARN_ON_ONCE pattern as the TX stats bounds fixes.
Add ath12k_wifi7_dp_rx_wbm_err_free_skb() helper to combine drop stat increment and skb free at each drop site, reducing repetition. Define the helper at the top of dp_rx.c before its first use.
Increment WBM_ERR_DROP_RXDMA_GENERIC only inside rxdma_err()'s
default: branch so it counts only unhandled RXDMA errors, not every RXDMA drop - consistent with how WBM_ERR_DROP_REO_GENERIC is counted.
Expose WBM RX drop counts in the device_dp_stats debugfs file under a new 'WBM Rx Drop Count' section.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/debugfs.c | 23 +++++++++- drivers/net/wireless/ath/ath12k/dp.h | 10 +++- drivers/net/wireless/ath/ath12k/dp_stats.h | 25 ++++++++++ drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c | 46 +++++++++++++++---- .../net/wireless/ath/ath12k/wifi7/hal_rx.c | 5 +- 5 files changed, 95 insertions(+), 14 deletions(-) create mode 100644 drivers/net/wireless/ath/ath12k/dp_stats.h
drivers/net/wireless/ath/ath12k/debugfs.c +21 -2
@@ -1071,6 +1071,20 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file, [HAL_REO_DEST_RING_ERROR_CODE_PN_ERR_FLAG_SET] = "PN err", [HAL_REO_DEST_RING_ERROR_CODE_DESC_BLOCKED] = "Desc blocked"}; +static const char *wbm_rx_drop[WBM_ERR_DROP_MAX ] = { + [WBM_ERR_GET_SW_DESC ] ="SW desc error" , + [WBM_ERR_DESC_PARSE ] ="Desc parse error" , + [WBM_ERR_DROP_INV_HW_ID ] ="Invalid hw id" , + [WBM_ERR_DROP_NULL_PRTNR_DP ] ="Null Partner dp" , + [WBM_ERR_DROP_NULL_PROC_DP ] ="Process Null Partner dp" , + [WBM_ERR_DROP_NULL_PDEV ] ="Null Pdev" , + [WBM_ERR_DROP_NULL_AR ] ="Null ar" , + [WBM_ERR_DROP_CAC_RUNNING ] ="CAC Running" , + [WBM_ERR_DROP_SG ] ="Scatter Gather" , + [WBM_ERR_DROP_INV_NWIFI_HDR ] ="Invalid NWifi Hdr len" , + [WBM_ERR_DROP_REO_GENERIC ] ="REO Generic" , + [WBM_ERR_DROP_RXDMA_GENERIC ] ="RXDMA Generic" }; +static const char *wbm_rel_src[HAL_WBM_REL_SRC_MODULE_MAX ] = { [HAL_WBM_REL_SRC_MODULE_TQM ] ="TQM" , [HAL_WBM_REL_SRC_MODULE_RXDMA ] ="Rxdma" ,
@@ -1095,13 +1109,18 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file,for (i =0 ; i <HAL_REO_ENTR_RING_RXDMA_ECODE_MAX ; i++) len +=scnprintf (buf + len, size - len,"%s: %u\n" , - rxdma_err[i], device_stats->rxdma_error[i]); + rxdma_err[i], device_stats->wbm_err.rxdma_error[i]); len +=scnprintf (buf + len, size - len,"\nREO errors:\n" );for (i =0 ; i <HAL_REO_DEST_RING_ERROR_CODE_MAX ; i++) len +=scnprintf (buf + len, size - len,"%s: %u\n" , - reo_err[i], device_stats->reo_error[i]); + reo_err[i], device_stats->wbm_err.reo_error[i]); + + len +=scnprintf (buf + len, size - len,"\nWBM Rx Drop Count:\n" ); +for (i =0 ; i <WBM_ERR_DROP_MAX ; i++) + len +=scnprintf (buf + len, size - len,"%s: %u\n" , + wbm_rx_drop[i], device_stats->wbm_err.drop[i]); len +=scnprintf (buf + len, size - len,"\nHAL REO errors:\n" );
drivers/net/wireless/ath/ath12k/dp.h +8 -2
@@ -7,6 +7,7 @@#ifndef ATH12K_DP_H #define ATH12K_DP_H +#include "dp_stats.h" #include "hw.h" #include "dp_htt.h" #include "dp_cmn.h"
@@ -436,13 +437,18 @@ struct ath12k_device_dp_tx_err_stats {atomic_t misc_fail; }; +struct ath12k_device_dp_rx_wbm_err_stats { +u32 rxdma_error[HAL_REO_ENTR_RING_RXDMA_ECODE_MAX ]; +u32 reo_error[HAL_REO_DEST_RING_ERROR_CODE_MAX ]; +u32 drop[WBM_ERR_DROP_MAX ]; +}; +struct ath12k_device_dp_stats {u32 err_ring_pkts;u32 invalid_rbm; -u32 rxdma_error[HAL_REO_ENTR_RING_RXDMA_ECODE_MAX ]; -u32 reo_error[HAL_REO_DEST_RING_ERROR_CODE_MAX ];u32 hal_reo_error[DP_REO_DST_RING_MAX ];struct ath12k_device_dp_tx_err_stats tx_err; +struct ath12k_device_dp_rx_wbm_err_stats wbm_err;u32 reo_rx[DP_REO_DST_RING_MAX ][ATH12K_MAX_DEVICES ];u32 rx_wbm_rel_source[HAL_WBM_REL_SRC_MODULE_MAX ][ATH12K_MAX_DEVICES ];u32 tqm_rel_reason[MAX_TQM_RELEASE_REASON ];
drivers/net/wireless/ath/ath12k/dp_stats.h new file +25 -0
@@ -0,0 +1,25 @@ +/* SPDX-License-Identifier: BSD-3-Clause-Clear */ +/* +* Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries. +*/ + +#ifndef ATH12K_DP_STATS_H +#define ATH12K_DP_STATS_H + +enum ath12k_wbm_err_drop_reason { +WBM_ERR_GET_SW_DESC , +WBM_ERR_DESC_PARSE , +WBM_ERR_DROP_INV_HW_ID , +WBM_ERR_DROP_NULL_PRTNR_DP , +WBM_ERR_DROP_NULL_PROC_DP , +WBM_ERR_DROP_NULL_PDEV , +WBM_ERR_DROP_NULL_AR , +WBM_ERR_DROP_CAC_RUNNING , +WBM_ERR_DROP_SG , +WBM_ERR_DROP_INV_NWIFI_HDR , +WBM_ERR_DROP_REO_GENERIC , +WBM_ERR_DROP_RXDMA_GENERIC , +WBM_ERR_DROP_MAX , +}; + +#endif /* ATH12K_DP_STATS_H */
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +37 -9
134 lines not on this page · raw message
drivers/net/wireless/ath/ath12k/wifi7/hal_rx.c +4 -1
@@ -329,7 +329,10 @@ int ath12k_wifi7_hal_desc_reo_parse_err(struct ath12k_dp *dp, HAL_REO_DEST_RING_INFO0_PUSH_REASON); err_code =le32_get_bits (desc->info0,HAL_REO_DEST_RING_INFO0_ERROR_CODE ); - dp->device_stats.reo_error[err_code]++; +if (likely (err_code <HAL_REO_DEST_RING_ERROR_CODE_MAX )) + dp->device_stats.wbm_err.reo_error[err_code]++; +else +WARN_ON_ONCE (1 );if (push_reason !=HAL_REO_DEST_RING_PUSH_REASON_ERR_DETECTED && push_reason !=HAL_REO_DEST_RING_PUSH_REASON_ROUTING_INSTRUCTION ) {
Signature · 1 line
2.34.1
Re: [PATCH ath-next 5/8] wifi: ath12k: add WBM RX error drop statistics
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c int ath12k_wifi7_dp_rx_process_wbm_err(struct ath12k_dp *dp, · line 1990
Quoted 10 lines · Pardeep Kaur
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-1@@ -1961,7 +1983,8 @@ int ath12k_wifi7_dp_rx_process_wbm_err(struct ath12k_dp *dp, device_id = desc_info->device_id; partner_dp =ath12k_dp_hw_grp_to_dp (dp_hw_grp, device_id);if (unlikely (!partner_dp)) { -dev_kfree_skb_any (msdu); +ath12k_wifi7_dp_rx_wbm_err_free_skb (dp, msdu, +WBM_ERR_DROP_NULL_PRTNR_DP );/* In any case continuation bit is set * in the previous record, cleanup scatter_msdu_list
My review agent flagged that between the blob above and the blob below is the following code that drops an skb without accounting:
if (err_info.push_reason !=
HAL_REO_DEST_RING_PUSH_REASON_ERR_DETECTED) {
dev_kfree_skb_any(msdu);
continue;
}Should this be counted?
Quoted 10 lines
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-1@@ -2007,7 +2030,8 @@ int ath12k_wifi7_dp_rx_process_wbm_err(struct ath12k_dp *dp, hw_link_id =ath12k_dp_rx_get_msdu_src_link (partner_dp->hal, msdu_data);if (hw_link_id >=ATH12K_GROUP_MAX_RADIO ) { -dev_kfree_skb_any (msdu); +ath12k_wifi7_dp_rx_wbm_err_free_skb (dp, msdu, +WBM_ERR_DROP_INV_HW_ID );/* In any case continuation bit is set * in the previous record, cleanup scatter_msdu_list
Re: [PATCH ath-next 5/8] wifi: ath12k: add WBM RX error drop statistics
Quoted 14 lines · Jeff Johnson
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c int ath12k_wifi7_dp_rx_process_wbm_err(struct ath12k_dp *dp, · line 1990
Quoted 10 lines · Pardeep Kaur
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-1@@ -1961,7 +1983,8 @@ int ath12k_wifi7_dp_rx_process_wbm_err(struct ath12k_dp *dp, device_id = desc_info->device_id; partner_dp =ath12k_dp_hw_grp_to_dp (dp_hw_grp, device_id);if (unlikely (!partner_dp)) { -dev_kfree_skb_any (msdu); +ath12k_wifi7_dp_rx_wbm_err_free_skb (dp, msdu, +WBM_ERR_DROP_NULL_PRTNR_DP );/* In any case continuation bit is set * in the previous record, cleanup scatter_msdu_list My review agent flagged that between the blob above and the blob below is the following code that drops an skb without accounting:
Quoted 7 lines
if (err_info.push_reason != HAL_REO_DEST_RING_PUSH_REASON_ERR_DETECTED) { dev_kfree_skb_any(msdu); continue; }Should this be counted?
This one was not handled as part of the current patch incremental patch will handle this
Quoted 11 lines
Quoted 11 lines
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-1@@ -2007,7 +2030,8 @@ int ath12k_wifi7_dp_rx_process_wbm_err(struct ath12k_dp *dp, hw_link_id =ath12k_dp_rx_get_msdu_src_link (partner_dp->hal, msdu_data);if (hw_link_id >=ATH12K_GROUP_MAX_RADIO ) { -dev_kfree_skb_any (msdu); +ath12k_wifi7_dp_rx_wbm_err_free_skb (dp, msdu, +WBM_ERR_DROP_INV_HW_ID );/* In any case continuation bit is set * in the previous record, cleanup scatter_msdu_list
PATCH 6/8 wifi: ath12k: track per-ring RX sent-to-stack count
In reply to Pardeep KaurFrom: Hariharan Ramanathan <[email protected]>
The device DP stats provide visibility into RX errors and drops but lack a counter for successfully delivered packets. Without this, it is difficult to correlate REO ring activity with actual stack delivery during debugging or performance analysis.
Add sent_to_stack[DP_REO_DST_RING_MAX] to ath12k_device_dp_stats to track successful deliveries per REO ring. A per-device dimension is not needed because partner_dp is already resolved at the increment point, so the counter is incremented directly in partner_dp->device_stats, attributing it to the correct device without an extra index.
Increment the counter just before each MSDU is handed off to the stack via ath12k_dp_rx_deliver_msdu(). Add a likely()/WARN_ON_ONCE() bounds check on ring_id before indexing sent_to_stack[], consistent with all other new stat arrays in this series.
Expose the per-ring counts in the device_dp_stats debugfs file under a new 'REO sent to stack' section.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/debugfs.c | 5 +++++ drivers/net/wireless/ath/ath12k/dp.h | 1 + drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c | 5 ++++- 3 files changed, 10 insertions(+), 1 deletion(-)
drivers/net/wireless/ath/ath12k/debugfs.c +5 -0
@@ -1122,6 +1122,11 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file, len +=scnprintf (buf + len, size - len,"%s: %u\n" , wbm_rx_drop[i], device_stats->wbm_err.drop[i]); + len +=scnprintf (buf + len, size - len,"\nREO sent to stack\n" ); +for (i =0 ; i <DP_REO_DST_RING_MAX ; i++) + len +=scnprintf (buf + len, size - len,"ring%d: %u\n" , + i, device_stats->sent_to_stack[i]); + len +=scnprintf (buf + len, size - len,"\nHAL REO errors:\n" );for (i =0 ; i <DP_REO_DST_RING_MAX ; i++)
drivers/net/wireless/ath/ath12k/dp.h +1 -0
@@ -457,6 +457,7 @@ struct ath12k_device_dp_stats {u32 tx_enqueued[DP_TCL_NUM_RING_MAX ];u32 tx_completed[DP_TCL_NUM_RING_MAX ];u32 reo_excep_msdu_buf_type; +u32 sent_to_stack[DP_REO_DST_RING_MAX ]; };struct ath12k_dp {
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +4 -1
@@ -651,7 +651,10 @@ ath12k_wifi7_dp_rx_process_received_packets(struct ath12k_dp *dp,dev_kfree_skb_any (msdu);continue ; } - +if (likely (ring_id <DP_REO_DST_RING_MAX )) + partner_dp->device_stats.sent_to_stack[ring_id]++; +else +WARN_ON_ONCE (1 );ath12k_dp_rx_deliver_msdu (dp_pdev, napi, msdu, &rx_info); }
Signature · 1 line
2.34.1
Re: [PATCH ath-next 6/8] wifi: ath12k: track per-ring RX sent-to-stack count
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c ath12k_wifi7_dp_rx_process_received_packets(struct ath12k_dp *dp, · line 655
Quoted 11 lines · Pardeep Kaur
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-1@@ -651,7 +651,10 @@ ath12k_wifi7_dp_rx_process_received_packets(struct ath12k_dp *dp,dev_kfree_skb_any (msdu);continue ; } - +if (likely (ring_id <DP_REO_DST_RING_MAX )) + partner_dp->device_stats.sent_to_stack[ring_id]++;
My review agent had the following finding which I did not try to validate:
sent_to_stack[] is incremented on partner_dp but reo_rx[] is on dp — this splits correlated stats across two devices' debugfs files in MLO
In a 2-device MLO group, ath12k_wifi7_dp_rx_process_received_packets() increments partner_dp->device_stats.sent_to_stack[ring_id] (the msdu's originating device), while ath12k_wifi7_dp_rx_process() increments dp->device_stats.reo_rx[ring_id][device_id] (the ring owner).
Reading /sys/kernel/debug/.../device0/device_dp_stats shows reo_rx[ring0]=N but sent_to_stack[ring0]=0; device1's file shows the reverse. The two counters meant to track the same traffic path cannot be compared from a single file.
Quoted 5 lines
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-0+else +WARN_ON_ONCE (1 );ath12k_dp_rx_deliver_msdu (dp_pdev, napi, msdu, &rx_info); }
Re: [PATCH ath-next 6/8] wifi: ath12k: track per-ring RX sent-to-stack count
Quoted 17 lines · Jeff Johnson
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c ath12k_wifi7_dp_rx_process_received_packets(struct ath12k_dp *dp, · line 655
Quoted 11 lines · Pardeep Kaur
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-1@@ -651,7 +651,10 @@ ath12k_wifi7_dp_rx_process_received_packets(struct ath12k_dp *dp,dev_kfree_skb_any (msdu);continue ; } - +if (likely (ring_id <DP_REO_DST_RING_MAX )) + partner_dp->device_stats.sent_to_stack[ring_id]++;My review agent had the following finding which I did not try to validate:
sent_to_stack[] is incremented on
partner_dpbut reo_rx[] is on dp — this splits correlated stats across two devices' debugfs files in MLO
Quoted 8 lines
In a 2-device MLO group,
ath12k_wifi7_dp_rx_process_received_packets()increments partner_dp->device_stats.sent_to_stack[ring_id] (the msdu's originating device), whileath12k_wifi7_dp_rx_process()increments dp->device_stats.reo_rx[ring_id][device_id] (the ring owner).Reading
/sys/kernel/debug/.../device0/device_dp_statsshows reo_rx[ring0]=N but sent_to_stack[ring0]=0; device1's file shows the reverse. The two counters meant to track the same traffic path cannot be compared from a single file.
will handle this with adding the device_id and ring id
Quoted 7 lines
Quoted 7 lines
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +2
-0+else +WARN_ON_ONCE (1 );ath12k_dp_rx_deliver_msdu (dp_pdev, napi, msdu, &rx_info); }
PATCH 7/8 wifi: ath12k: fix 1-based ring index in REO Rx Received debugfs output
In reply to Pardeep KaurFrom: Pardeep Kaur <[email protected]>
The REO Rx Received section in device_dp_stats debugfs uses a 1-based ring index ("Ring1:", "Ring2:", ...) which is inconsistent with the rest of ath12k where rings are indexed from 0. Fix it to use 0-based indexing and lowercase "ring%d:" to match the style used by all other ring counters in the same function.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/debugfs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/wireless/ath/ath12k/debugfs.c +1 -1
@@ -1210,7 +1210,7 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file, len +=scnprintf (buf + len, size - len,"\nREO Rx Received:\n" );for (i =0 ; i <DP_REO_DST_RING_MAX ; i++) { - len +=scnprintf (buf + len, size - len,"Ring%d:" , i +1 ); + len +=scnprintf (buf + len, size - len,"ring%d:" , i);for (j =0 ; j < ab->ag->num_devices; j++) { len += scnprintf(buf + len, size - len,
Signature · 1 line
2.34.1
PATCH 8/8 wifi: ath12k: add WBM SW desc fallback counter
In reply to Pardeep KaurFrom: Aniruddha Mishra <[email protected]>
When HW CC is not done, the WBM RX error path falls back to cookie-based descriptor retrieval via ath12k_dp_get_rx_desc(). This fallback path is taken before knowing whether the retrieval will succeed, so it is not a drop event and should not be counted in the drop[] array.
Add a dedicated sw_desc_fallback counter in ath12k_device_dp_rx_wbm_err_stats to track how often the HW CC fallback path is taken, distinct from actual packet drops. Expose the counter in the device_dp_stats debugfs file as a separate line.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.6.r1-00402-QCAHKSWPL_SILICONZ-1
--- drivers/net/wireless/ath/ath12k/debugfs.c | 4 ++++ drivers/net/wireless/ath/ath12k/dp.h | 1 + drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c | 1 + 3 files changed, 6 insertions(+)
drivers/net/wireless/ath/ath12k/debugfs.c +4 -0
@@ -1122,6 +1122,10 @@ static ssize_t ath12k_debugfs_dump_device_dp_stats(struct file *file, len +=scnprintf (buf + len, size - len,"%s: %u\n" , wbm_rx_drop[i], device_stats->wbm_err.drop[i]); + len +=scnprintf (buf + len, size - len, +"\nWBM SW desc fallback (HW CC not done): %u\n" , + device_stats->wbm_err.sw_desc_fallback); + len +=scnprintf (buf + len, size - len,"\nREO sent to stack\n" );for (i =0 ; i <DP_REO_DST_RING_MAX ; i++) len += scnprintf(buf + len, size - len, "ring%d: %u\n",
drivers/net/wireless/ath/ath12k/dp.h +1 -0
@@ -441,6 +441,7 @@ struct ath12k_device_dp_rx_wbm_err_stats {u32 rxdma_error[HAL_REO_ENTR_RING_RXDMA_ECODE_MAX ];u32 reo_error[HAL_REO_DEST_RING_ERROR_CODE_MAX ];u32 drop[WBM_ERR_DROP_MAX ]; +u32 sw_desc_fallback; };struct ath12k_device_dp_stats {
drivers/net/wireless/ath/ath12k/wifi7/dp_rx.c +1 -0
@@ -1968,6 +1968,7 @@ int ath12k_wifi7_dp_rx_process_wbm_err(struct ath12k_dp *dp,/* retry manual desc retrieval if hw cc is not done */ if (!desc_info) { + dp->device_stats.wbm_err.sw_desc_fallback++; desc_info =ath12k_dp_get_rx_desc (dp, err_info.cookie);if (!desc_info) { dp->device_stats.wbm_err.drop[WBM_ERR_GET_SW_DESC ]++;
Signature · 1 line
2.34.1