In the function ath12k_htt_mlo_offset_event_handler(), it is possible that
the ar is not yet active (started). The function
ath12k_mac_get_ar_by_pdev_id() only searches for active pdev, so a NULL
return is possible. Therefore, there is no need to print a warning,
instead, just silently discard the message.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Aditya Kumar Singh <quic_adisi@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241211154358.776279-5-kvalo@kernel.org
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently, partner device buffer is not handled in the receive data path. In
Multi-Link Operation, the partner device buffer is reported to the primary
upper MAC rings. Therefore, add partner device buffer processing in the REO and
Exception ring handler.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241209185421.376381-3-kvalo@kernel.org
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently, all the rx processing treats the hardware link id as the MAC id.
The HW link id is a unique identifier for all hardware links participating in
the multi link group. Therefore, the current MAC id derivation is insufficient
to process the partner rx buffer in the Multi-Link Operation. So derive the
MAC id from the HW link id and implementing this change in rx processing will
enable the scaling of partner buffer processing in Multi-Link Operation.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241209185421.376381-2-kvalo@kernel.org
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently, in DP rx error processing, the MAC id is fetched redundantly from
the same descriptor for each MSDU. To avoid this redundancy, move the fetch
handling before the iteration.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241211153432.775335-9-kvalo@kernel.org
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
In ath12k_dp_rx_peer_frag_setup(), commit ea41925538 ("wifi: ath12k: add primary
link for data path operations") checks whether a link is the primary link, and
returns directly if it isn't. In ML scenario where we have non-primary links created,
this results in leaking the michael_mic info since it is allocated by default but
could never be freed for a non-primary link.
Note that we can not move the might-sleep allocation after primary link check since
there we are in atomic context (due to spin lock). So keep the default allocation,
and then free it before return to fix this issue.
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4
Fixes: ea41925538 ("wifi: ath12k: add primary link for data path operations")
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Acked-by: Kalle Valo <kvalo@kernel.org>
Acked-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Link: https://patch.msgid.link/20241206054552.177424-1-quic_bqiang@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Apply tid queue setup based on all link stations on receiving ampdu action
params for an ML Station.
Modify ath12k_get_ar_by_vif() to fetch ar based on link arvif inside ahvif.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Sriram R <quic_srirrama@quicinc.com>
Signed-off-by: Rameshkumar Sundaram <quic_ramess@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241126171139.2350704-5-kvalo@kernel.org
Currently reoqueue tid setup uses lookup table (LUT) during peer association,
but for ML peer there will be multiple link peers (belonging to different
underlying firmware) affiliated to each other. Hence the reo queue should be
setup only on one of the links which is the primary link.
Add changes to create separate ML reo queue lookup table for ML peers and use
the same while setting up rx tid for ML peer's primary link. For ML peers use
ml_peer_id instead of peer_id to setup/lookup the reo queue entry in the LUT.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Rameshkumar Sundaram <quic_ramess@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241121155806.1862733-6-kvalo@kernel.org
In case of Multi-link operation, data path peer setup and tid setup should be
done only for primary link of multi-link station. Add changes to introduce
primary link is peer. Currently, association link will be considered as primary
link.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Sriram R <quic_srirrama@quicinc.com>
Signed-off-by: Harshitha Prem <quic_hprem@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241121155806.1862733-4-kvalo@kernel.org
To prepare the driver for MLO support, split the driver sta data structure to
scale for multiple links. This requires changing the use of arsta to per link
and not per hw which can now comprise of multiple links. Also since most
configurations from mac80211 are done per link, do refactoring of the driver
functions to apply these configurations at link level.
Split struct ath12k_sta, which is the driver private of struct ieee80211_sta,
to store link specific information as struct ath12k_link_sta. For default use
cases struct ath12k_sta will have a preallocated link sta called deflink which
will be used by non ML STAs and the first link sta of ML STA.
With MLO support to be added, remaining link stas will allocated during state
change where new STA is added. These link stas will be freed when STA is
deleted.
Current ath12k_sta(arsta) structure:
+-----------------+ +----------------+
| | | |
| ieee80211_sta | | ieee80211_sta |
| private data | | private data |
| | | |
| ath12k_sta | | ath12k_sta |
| (arsta) | | (arsta) |
|+---------------+| | +-------------+|
|| *arvif (link || | |*arvif (link ||
|| vif of an ar || | | vif of an ar||
|| say 5GHz) || | | say 6GHz) ||
|+---------------+| | +-------------+|
+-----------------+ +----------------+
New struct ath12k_sta (ahsta) containing ath12k_link_sta(s) (arsta) (deflink is
preallocated member which is always the first link if ieee80211_sta is ML STA
and is the only link sta otherwise):
+---------------------------------------+
| ieee80211_sta |
| private data |
| |
| ath12k_sta (ahsta) |
| +-------------------------------------+
| | ath12k_link_sta deflink (arsta) |
| | |
| | *arvif (link vif of ar (5GHz)) |
| +-------------------------------------+
| +-------------------------------------+
| | ath12k_link_sta *link (arsta) |
| | |
| | *arvif (link vif of ar (6GHz)) |
| | |
| +-------------------------------------+
| |
+---------------------------------------+
To refactor existing ath12k_sta to make use of link stas, following changes are
made:
1. Limit struct ieee80211_sta argument mac80211 ops unless otherwise really required.
2. struct ath12k_sta, now called by variable name arsta, stores multiple
arstas (struct ah12k_link_sta) and also has a back pointer to struct
ath12k_sta.
3. Pass struct ath12k_link_sta to mac functions that passed struct
ieee80211_sta argument and fetch struct ath12k_sta (ahsta) and struct
ieee80211_sta(sta) internally. This is done to avoid passing link id in all
the functions and performing validation across these functions. Rather the
validation and sta to arsta conversion can be done only at the mac80211 ops.
4. In this patchset, only ahsta->deflink is used to be on par with the
existing code. When MLO support is added the link id will be used to fetch the
arsta.
5. Change ath12k_sta_to_arsta() to ath12k_vif_to_ahsta() to fetch the
ML level sta. The link sta can be fetched from ahsta->link[], or the
deflink can be accessed via ahsta->deflink. API to access link
sta (arsta) by passing link_id can be introduced with MLO Support.
6. The ieee80211_sta can be accessed from ahsta using ath12k_ahsta_to_sta()
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Sriram R <quic_srirrama@quicinc.com>
Co-developed-by: Rameshkumar Sundaram <quic_ramess@quicinc.com>
Signed-off-by: Rameshkumar Sundaram <quic_ramess@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241015171416.518022-4-kvalo@kernel.org
Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
To prepare the driver for MLO support, split the driver vif data structure to
scale for multiple links. This requires changing the use of arvif to per link
and not per hw which can now comprise of multiple links. Also since most
configurations from mac80211 are done per link, do refactoring of the driver
functions to apply these configurations at link level.
Split ath12k_vif which is the driver private of ieee80211_vif to store link
specific information as ath12k_link_vif. For default use cases the ath12k vif
will have a preallocated link vif called deflink which will be used by non ML
and the first link vif of ML vif.
With MLO support to be added, remaining link vifs will be allocated during
channel assignment where vdev create/start happens. These link vifs will be
freed during interface down.
Current ath12k_vif(arvif) structure
+---------------+ +---------------+ +---------------+
| ieee80211_vif | | ieee80211_vif | | ieee80211_vif |
| private data | | private data | | private data |
| | | | | |
| ath12k_vif | | ath12k_vif | | ath12k_vif |
| (arvif) | | (arvif) | | (arvif) |
| | | | | |
| +----------+ | | +----------+ | | +----------+ |
| |*ar (2GHz)| | | |*ar (5GHz)| | | |*ar (2GHz)| |
| +----------+ | | +----------+ | | +----------+ |
| | | | | |
+---------------+ +---------------+ +---------------+
The new ath12k_vif (ahvif) containing ath12k_link_vif(s) (arvif) (deflink is
preallocated member which is always the first link if ieee80211_vif is MLD and
is the only link otherwise):
+---------------------------------+
| ieee80211_vif |
| private data |
| |
| ath12k_vif(ahvif) |
| |
| +-------------------------------+
| |ath12k_link_vif deflink (arvif)|
| | +---------------+ |
| | | *ar(2GHz) | |
| +-------------------------------+
| +-------------------------------+
| | ath12k_link_vif *link (arvif)|
| | +---------------+ |
| | | *ar(5GHz) | |
| +-------------------------------+
| +-------------------------------+
| | ath12k_link_vif *link (arvif)|
| | +---------------+ |
| | | *ar(6GHz) | |
| +-------------------------------+
| |
+---------------------------------+
To refactor existing ath12k_vif to make use of link vifs, following
changes are made:
1. ath12k_vif now called by variable name ahvif storing multiple
arvifs (struct ah12k_link_vif) and also has a back pointer to ieee80211_vif.
2. In this patch set, only deflink is used to be on par with the
existing code. When MLO support is added the link id will be used to fetch
the arvif.
3. For mac80211 ops which doesn't use specific link_id, the config or info
is common for the vif, hence apply the config to all link vifs.
The links_map in the ahvif, will be used to identify all the link vifs that
are setup.
4. Change ath12k_vif_to_arvif() as ath12k_vif_to_ahvif() to fetch the
hw level vif. The link vif can be fetched from ahvif->link[], or the
deflink can be accessed via ahvif->deflink. API to access link
vif (arvif) by passing link_id can be introduced with MLO Support.
5. The ieee80211_vif can be accessed from ahvif using ath12k_ahvif_to_vif().
The locking continues to use wiphy_lock() for protecting access to most data in
struct ath12k&co, there are no changes in that regard. Though struct
ath12k_vif::link[] is protected with RCU.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Sriram R <quic_srirrama@quicinc.com>
Co-developed-by: Rameshkumar Sundaram <quic_ramess@quicinc.com>
Signed-off-by: Rameshkumar Sundaram <quic_ramess@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20241015171416.518022-2-kvalo@kernel.org
Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Currently, the ath12k_soc_dp_stats::hal_reo_error array is defined with a
maximum size of DP_REO_DST_RING_MAX. However, the ath12k_dp_rx_process()
function access ath12k_soc_dp_stats::hal_reo_error using the REO
destination SRNG ring ID, which is incorrect. SRNG ring ID differ from
normal ring ID, and this usage leads to out-of-bounds array access. To
fix this issue, modify ath12k_dp_rx_process() to use the normal ring ID
directly instead of the SRNG ring ID to avoid out-of-bounds array access.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20240704070811.4186543-2-quic_periyasa@quicinc.com
Currently, the Rx data path only supports parsing peer metadata of version
zero. However, the QCN9274 platform configures the peer metadata version
as V1B. When V1B peer metadata is parsed using the version zero logic,
invalid data is populated, causing valid packets to be dropped. To address
this issue, refactor the peer metadata version and add the version based
parsing to populate the data from peer metadata correctly.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Fixes: 2870338109 ("wifi: ath12k: add support for peer meta data version")
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://patch.msgid.link/20240624145418.2043461-1-quic_periyasa@quicinc.com
Currently, in the Rx error processing handler, once an MSDU drop is
detected, the subsequent MSDUs get unintentionally dropped due to the
previous drop flag being retained across all MSDU processing, leading
to the discarding of valid MSDUs. To resolve this issue, the drop flag
should be reset to false before processing each descriptor.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240604062641.2956288-1-quic_periyasa@quicinc.com
When handling fragmented packets, the ath12k driver reassembles each
fragment into a normal packet and then reinjects it into the HW ring.
However, a firmware crash occurs during this reinjection process.
The issue arises because the driver populates peer metadata in
reo_ent_ring->queue_addr_lo, while the firmware expects the physical
address obtained from the corresponding peer’s queue descriptor. Fix it
by filling peer's queue descriptor's physical address in queue_addr_lo.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00209-QCAHKSWPL_SILICONZ-1
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Signed-off-by: P Praneesh <quic_ppranees@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240520070045.631029-4-quic_ppranees@quicinc.com
For fragmented packets, ath12k reassembles each fragment as a normal
packet and then reinjects it into HW ring. In this case, the DMA
direction should be DMA_TO_DEVICE, not DMA_FROM_DEVICE. Otherwise,
an invalid payload may be reinjected into the HW and
subsequently delivered to the host.
Given that arbitrary memory can be allocated to the skb buffer,
knowledge about the data contained in the reinjected buffer is lacking.
Consequently, there’s a risk of private information being leaked.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00209-QCAHKSWPL_SILICONZ-1
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Co-developed-by: Baochen Qiang <quic_bqiang@quicinc.com>
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Signed-off-by: P Praneesh <quic_ppranees@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240520070045.631029-2-quic_ppranees@quicinc.com
The ath12k_dp_get_rx_desc() failure log currently contains the same
information across multiple Rx data paths and lacks sufficient detail for
debugging purposes. To address this, change the ath12k_dp_get_rx_desc()
failure log to include cookie information along with Rx path details.
This will provide more specific data for debugging purposes.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 HW2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240516000807.1704913-3-quic_periyasa@quicinc.com
The frequency in the rx status is currently being filled
incorrectly for the 6 GHz band. The channel number received is
invalid in this case, resulting in packet drops. Fix this
issue by correcting the frequency calculation.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Signed-off-by: Pradeep Kumar Chitrapu <quic_pradeepc@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240508173655.22191-3-quic_pradeepc@quicinc.com
smatch flagged the following issue:
drivers/net/wireless/ath/ath12k/dp_rx.c:4065 ath12k_dp_rxdma_ring_sel_config_wcn7850() error: uninitialized symbol 'ret'.
In ath12k_dp_rxdma_ring_sel_config_wcn7850() if it were ever the case
that ab->hw_params->num_rxdma_per_pdev was 0 then 'ret' would be
uninitialized when it is returned. This should never be the case, but
to be safe and to quiet smatch, add an initializer to the declaration
of 'ret'.
No functional changes, compile tested only.
Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240504-ath12k_dp_rxdma_ring_sel_config_wcn7850-ret-v1-2-44d2843a2857@quicinc.com
Currently, the WBM idle ring descriptor setup uses implicit value
HAL_RX_BUF_RBM_WBM_DEV0_IDLE_DESC_LIST for the return buffer manager
parameter. To support inter-device MLO (Multi-link operation), this
parameter needs to be configure dynamically based on the device identifier
within the MLO group. Therefore, introduce a new argument to the helper
function ath12k_hal_set_link_desc_addr().
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240430165811.1377182-3-quic_periyasa@quicinc.com
Currently pktlog is stopped in suspend callback and started in
resume callback, and in either scenarios it's basically to
delete/modify ab->mon_reap_timer and to purge related rings.
For WCN7850 it's pointless because pktlog is not enabled: both
ab->mon_reap_timer and those rings are not initialized.
So remove pktlog handling in suspend/resume callbacks. And
further, remove these two functions and related callee because
no one is calling them.
Other chips are not affected because now only WCN7850 supports
suspend/resume.
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240412060620.27519-8-quic_bqiang@quicinc.com
Currently, when debug logs are enabled, messages such as the following
are found in the kernel log:
event mgmt rx skb pK-error len 209 ftype 00 stype 50
The "pK-error" comes from using %pK to display an skb address in an
interrupt context.
Per the guidance in the current "printk-formats" documentation, %pK
"is only intended when producing content of a file read by userspace
from e.g. procfs or sysfs, not for dmesg."
So replace %pK with %p in all of the ath12k_dbg() format strings.
Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240416-pk-v1-1-2c8852743e6d@quicinc.com
Currently, in the rx error data path, mac id is extracted from the
last 64bits of MSDU end description tag for each entry received in
the WBM error ring. Then, each entry will be updated into the MSDU
list for further processing. The extracted mac id is valid when a
single MSDU is not fragmented and received in the WBM error ring.
In scenarios where the size of a single MSDU received exceeds the
descriptor buffer size, resulting in fragmented or spillover MSDU
entries into the WBM error ring. In this case, the extracted mac id
from each spillover entry is invalid except the last spillover entry
of the MSDU. This invalid mac id leads to packet rejection.
To address this issue, check if the MSDU continuation flag is set,
then extract the valid mac id from the last spillover entry.
Propagate the valid mac id to all the spillover entries of the single
MSDU in the temporary MSDU list(scatter_msdu_list). Then, update this
into the MSDU list (msdu_list) for further processing.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Nithyanantham Paramasivam <quic_nithp@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240404141538.1277258-1-quic_nithp@quicinc.com
The Rxdma replenish handler currently contains redundant code in both the
success and error cases. To avoid the redundant code, implement a
conditional jump mechanism by placing a goto statement to prevent
unnecessary execution of the error handling code when the operation is
successful. This change streamline the code readability and enhance its
efficiency.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240320010615.91331-4-quic_periyasa@quicinc.com
When a packet arrives in Rx rings, the RX descriptor moves from the used
list to the free list. Then, the rxdma ring gets replenished, where the Rx
descriptor again moves from the free list to the used list. At the end, the
descriptor came to the used list with unnecessary list movement. The
descriptor used list is maintained in the Rxdma ring structure, which
creates lock contention for the list operations (add, delete) in the Rx
data path. Optimize the Rx data path by removing the used list from the
common Rxdma ring and maintain as a local variable in the Rx ring handler
itself, which avoid lock contention. Now, to find the used list descriptor
during descriptor cleanup, we need to check the in_use flag for each Rx
descriptor.
This is a simple UDP UL throughput test case results on x86+NUC device
with QCN9274 card, which clearly shows 8% to 12% improvement in the CPU
idle for the same ingress traffic.
Before:
Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle
Average: all 0.24 0.00 12.54 0.08 0.00 23.33 0.00 0.00 0.00 63.81
After:
Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle
Average: all 0.34 0.00 4.60 0.00 0.00 19.59 0.00 0.00 0.00 75.47
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240320010615.91331-3-quic_periyasa@quicinc.com
Currently, the Rxdma buffer replenish procedure implicitly retrieves the
number of free entries when the caller passes req_entries as zero. This is
helpful to replenishing the Rxdma ring when the caller is unaware of the
number of entries to replenish. After the Rxdma ring is initialized, we
can pass req_entries as zero to replenish all the free entry in the Rxdma
ring. This change helps to align the argument req_entries to the used list
entries for better code readability.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240320010615.91331-2-quic_periyasa@quicinc.com
When more than one pdev is supported in the chip/SoC, the packet
belonging to second pdev is given to first pdev due to not identifying
the MAC id in the WBM error path. So ping fails.
In WBM error path, src link id information not available in the
descriptor. So get this information from the msdu_end 64bit tag. It
is necessary to get the src link id to identify the MAC id in the
given chip. Then only we can pass the skb to the corresponding pdev.
The msdu_end 64bit tag is available only if compact Rx TLVs
descriptors are used. Thus, src link id is fetched correctly in WBM
error path when compact descriptors are in use.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00188-QCAHKSWPL_SILICONZ-1
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Raj Kumar Bhagat <quic_rajkbhag@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240129065724.2310207-11-quic_rajkbhag@quicinc.com
Most of the RX descriptors fields are currently not used in the
ath12k driver. Hence add support to selectively subscribe to the
required quad words (64 bits) within msdu_end and mpdu_start of
rx_desc.
Add compact rx_desc structures and configure the bit mask for Rx TLVs
(msdu_end, mpdu_start, mpdu_end) via registers. With these registers
SW can configure to DMA the partial TLV struct to Rx buffer.
Each TLV type has its own register to configure the mask value.
The mask value configured in register will indicate if a particular
QWORD has to be written to rx buffer or not i.e., if Nth bit is enabled
in the mask Nth QWORD will be written and it will not be written if the
bit is disabled in mask. While 0th bit indicates whether TLV tag will be
written or not.
Advantages of Qword subscription of TLVs
- Avoid multiple cache-line misses as the all the required fields
of the TLV are within 128 bytes.
- Memory optimization as TLVs + DATA + SHINFO can fit in 2k buffer
even for 64 bit kernel.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00188-QCAHKSWPL_SILICONZ-1
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Kathirvel <quic_kathirve@quicinc.com>
Co-developed-by: Raj Kumar Bhagat <quic_rajkbhag@quicinc.com>
Signed-off-by: Raj Kumar Bhagat <quic_rajkbhag@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240129065724.2310207-10-quic_rajkbhag@quicinc.com
With word mask subscription support, the rx_desc structure will
change. The fields in this structure rx_desc will be reduced to only
the required fields. To make word mask subscription changes compatible
with the older firmware version (firmware that does not support word
mask subscription), two different structures of rx_desc will be
required for the same hardware.
The hardware param hal_desc_sz value cannot be constant for the same
hardware. It depends on the size of rx_desc structure which may
change based on firmware capability to support word mask subscription.
Hence, remove hal_desc_sz from hardware param and add hal_rx_ops
to get the size of rx_desc in run time.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00188-QCAHKSWPL_SILICONZ-1
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Raj Kumar Bhagat <quic_rajkbhag@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240129065724.2310207-9-quic_rajkbhag@quicinc.com
With word mask subscription support, the rx_desc structure will
change. The fields in this structure rx_desc will be reduced to only
the required fields. To make word mask subscription changes compatible
with the older firmware version (firmware that does not support word
mask subscription), two different structures of rx_desc will be
required for the same hardware.
Most of the hal_ops are directly related to rx_desc.
Hence, split hal_ops into operations that are independent of rx_desc
(hal_ops) and the operations that are directly dependent on rx_desc
(hal_rx_ops). These hal_rx_ops depend on both hardware and firmware
hence move them out of hw_params.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00188-QCAHKSWPL_SILICONZ-1
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Raj Kumar Bhagat <quic_rajkbhag@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240129065724.2310207-8-quic_rajkbhag@quicinc.com
Currently, mac80211 hardware accesses link/radio structure directly in
multiple locations. Introduce helper function to avoid this direct
access, as this change will facilitate refactoring for Multi-link
operation support.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240118010320.3918136-2-quic_periyasa@quicinc.com
For WCN7850, there is a SRNG named MAC buffer ring, i.e.,
dp->rx_mac_buf_ring. During initialization, it is setup
by host and then under control of firmware. During RX
process, firmware fetches buffers from
dp->rx_refill_buf_ring to fill that MAC buffer ring,
and those buffers are taken by RXDMA to carry real WLAN
frames received from air.
Currently a low RX throughput is observed. Checking
firmware log, lots of errors are reported by MAC buffer
ring, complaining that it is running out of buffers,
which further indicates that RXDMA is suffering from
starvation. Currently the size of dp->rx_mac_buf_ring
is configured as 1024. After changing it to 2048, those
error messages are reduced, and a 6.4% increase is seen
in peak throughput. Note that 2048 is an empirical
value. It is chosen here because the RX throughput
meets our expectation after the change.
This change only applies to WCN7850 since other
chips don't have a MAC buffer ring.
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0-03427-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.15378.4
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20231129020414.56425-4-quic_bqiang@quicinc.com
Add 320 MHz bandwidth as a new bandwidth enum for ATH12k driver. This
is extending existing bandwidth related enums to include 320 MHz. This
is a precursor to supporting 320 MHz in the future.
Sanity test performed to confirm that there was no impact in existing
bandwidths.
Additionally update QuIC copyright to include 2023 in hal_rx.h.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00125-QCAHKSWPL_SILICONZ-1
Signed-off-by: Muna Sinada <quic_msinada@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231116221839.1303170-1-quic_msinada@quicinc.com
Currently data path Rxdma ring structure store the IDR buffer and lock.
These IDR handling is needed only for SW cookie conversion and not
needed for HW cookie conversion. REO Rxdma ring use the HW cookie
conversion and monitor Rxdma ring use the SW cookie conversion.
Since idr not needed for REO Rxdma ring, remove the IDR data entity
from the data path Rxdma ring structure. Introduce the new data path ring
structure for monitor rxmda rings since it need IDR data entity.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00125-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231111043934.20485-5-quic_periyasa@quicinc.com
Currently Rxdma replenish require HW conversion argument which is
unnecessary argument since ath12k driver configures the Rxdma only
in HW conversion. To optimize the rx data path per packet, avoid
the explicit unnecessary argument and condition check in the rx
replenish.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00125-QCAHKSWPL_SILICONZ-1
Signed-off-by: Karthikeyan Periyasamy <quic_periyasa@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231111043934.20485-4-quic_periyasa@quicinc.com
Currently, the logic to return an ath12k_sta pointer, given a
ieee80211_sta pointer, uses typecasting throughout the driver. In
general, conversion functions are preferable to typecasting since
using a conversion function allows the compiler to validate the types
of both the input and output parameters.
ath12k already defines a conversion function ath12k_vif_to_arvif() for
a similar conversion. So introduce ath12k_sta_to_arsta() for this use
case, and convert all of the existing typecasting to use this
function.
Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231019-upstream-ath12k_sta_to_arsta-v1-1-06f06f693338@quicinc.com
>From 'IEEE Std 802.11-2020 section 11.3.4.1':
If STA A in an infrastructure BSS receives a Class 2 or Class 3 frame
from STA B that is not authenticated with STA A
(i.e., the state for STA B is State 1), STA A shall discard the frame.
If the frame has an individual address in the Address 1 field,
the MLME of STA A shall send a Deauthentication frame to STA B.
When data frames from unassociated stations are received by an AP,
the AP is supposed to send a deauthentication/disassociation frame with
reason code "Class 2 frame received from nonauthenticated STA" or
"Class 3 frame received from nonassociated STA".
But ath12k AP doesn't send deauthentication/disassociation frames,
when it receives data frames from unassociated stations.
The ath12k driver drops the data frames from unassociated
station and the upper layer(mac80211/hostapd) is not aware of such event.
Hence deauthentication/disassociation frame is not sent to that
particular station by the AP.
To address this issue, allow the data frames from the
unassociated stations to reach mac80211 so that mac80211 can send
NL80211_CMD_UNEXPECTED_FRAME event to userspace(hostapd) and hostapd
upon receiving the event will send the deauthentication/disassociation
frame with proper reason code.
The data frame from unassociated stations gets dropped in mac80211.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0-02903-QCAHKSWPL_SILICONZ-1
Signed-off-by: Hari Chandrakanthan <quic_haric@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/1689749074-14676-1-git-send-email-quic_haric@quicinc.com
len is extracted from HTT message and could be an unexpected value in
case errors happen, so add validation before using to avoid possible
out-of-bound read in the following message iteration and parsing.
The same issue also applies to ppdu_info->ppdu_stats.common.num_users,
so validate it before using too.
These are found during code review.
Compile test only.
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20230901015602.45112-1-quic_bqiang@quicinc.com
