Increase ce_service_max_yield_time to 2 msec in SLUB_DEBUG
builds for helium SNOC platforms to improve the CE reaping
performance.
Change-Id: I2a8a21fd1a65bc2ffc5fb17986116613f47510c7
CRs-Fixed: 2937231
TWT notify event from firmware is expected only if the teardown
is initiated by the firmware during scenarios like
(roaming in progress, etc).
If the TWT teardown is initiated by the peer then notify WMI event
is not expected from the firmware.
Change-Id: Ic13a4858fb06d023d5e71da10320a546a1dcf80d
CRs-Fixed: 2944663
Sometimes we need PDEV suspend mode, WMI_PDEV_SUSPEND_CMDID and
WMI_PDEV_RESUME_CMDID. But currently design, only support WoW
suspend or not support two choices. And, there are two confused
ini configuration, gDisableWow and gEnableWoW.
1. Add new ini gSuspendMode to set suspend mode. Default 2.
0, does not support suspend
1, legency pdev suspend mode
2, WoW suspend mode
2. Clean gDisableWow. It is replaced by gSuspendMode=0.
HDD suspend/resume function get the value by new API
ucfg_pmo_get_suspend_mode().
Change-Id: Icbeef3df6879ca2054f64e0292e9b9469c4936d4
CRs-fixed: 2940788
The INI skip_tpe_consideration when set, skips power from AP's
TPE IE in TPC power calculation for 2.4G/5G case. Enable this
INI by default since APs advertise low TPE powers for IoT, thus
lowering overall TPC power.
Change-Id: I6c07bbd684579c2fa3617e49d76c75b5bf3f2815
CRs-Fixed: 2941630
Fragments are not flushed as part of rekey which
could result in fragments encrypted under different
keys to be reassembled.
Fix is to flush fragments for the peer for which add
key request is received.
Change-Id: I0c018ff7375272125c62aaea7b8ad4df9e842508
CRs-Fixed: 2875950
wma_send_peer_assoc() calls wma_unified_peer_state_update() which
always sets the peer state as AUTH irrespective of peer state.
Remove wma_unified_peer_state_update() from wma_send_peer_assoc()
to handle peer state.
Change-Id: I4a887acbc8018653c34b927636cce7cc05323838
CRs-Fixed: 2888808
For some of the stability issues triggered by H.W,
by the time host detects it most of the QDF history
is getting overwritten to avoid this increasing QDF
nbuf history. So that there will be less chance of
events getting overwritten.
Change-Id: Ia4bad1a175976af2e56126b56ae8c92a988e4289
CRs-Fixed: 2942868
Fragmented EAPOL frames and EAPOL frames received
in few error scenarios with DA different from SAP
vdev mac addr will be dropped.
Change-Id: I624eba5bdb43c6b88a1f57112550f8026cc35e24
CRs-Fixed: 2888227
Do not intrabss forward fragmented EAPOL frames that have
DA different from the SAP vdev mac addr when high latency
is enabled.
Change-Id: Idb6e6c001f4dae51c2181e70ab9adbbb964f0ee3
CRs-Fixed: 2942096
Drop non-EAPOL/WAPI frames from unauthorized peer received
in the IPA exception path.
Change-Id: I0c0bc6e60efa193126ba1e3eca36c5e02f7f76a3
CRs-Fixed: 2860206
Do not intrabss forward fragmented EAPOL frames that have
DA different from the SAP vdev mac addr.
Change-Id: I4145227c9b02fe8cec86ef4ffc3bc2025f906923
CRs-Fixed: 2888467
Modify check to ensure packet number is consecutive for
fragments and drop the fragments if the check fails.
Change-Id: Ica24f65aff65ca58bb010c876f27964b5b2bae6a
CRs-Fixed: 2860242
Multicast frames should not be fragmented and plaintext
frags should not be reassembeld in protected network.
Fix is to drop mcast frags and plaintext frags received
in protected network.
Change-Id: I98cf0715f5832f2f86f86b79dbdbc3a7c86dbfd0
CRs-Fixed: 2860245
Some APs/P2P-GOs send beacon with 20mhz and assoc resp with 80mhz
and after assoc resp, next beacon also has 80mhz. Connection is
expected to happen in better possible bandwidth(80MHz in this
case).
Start the vdev with max supported ch_width in order to
support this. It'll be downgraded to appropriate ch_width or
the same would be continued based on assoc resp.
Change-Id: I0d205e0833b2638aab15401bacdef7a0696d40c2
CRs-Fixed: 2870758
During DUT STA Rx over IPA, qdf_spin_unlock_bh is called as following.
__wlan_ipa_w2i_cb
->wlan_ipa_rx_intrabss_fwd
->cdp_ipa_rx_intrabss_fwd
->dp_ipa_rx_intrabss_fwd
->dp_vdev_get_ref_by_id
->qdf_spin_unlock_bh
netif_rx will raise NET_RX_SOFTIRQ in enqueue_to_backlog->
____napi_schedule even NAPI not enabled.
spin_unlock_bh will call do_softirq to execute NET_RX if NET_RX_SOFTIRQ
is raised.
If qdf_spin_unlock_bh is called in wlan_ipa_w2i_cb, once 1 skb received,
netif_rx is called, and NET_RX_SOFTIRQ and APP like iperf is scheduled,
too much context switch happens, CPU and power is wasted, Rx throughput
KPI failed on mdm platform which only has single 1.2G-1.5G CPU.
Fix it by avoiding intra bss forwarding for adapter in sta mode.
Change-Id: I264b14120b9998a56f6c1978f5a890615e6f837c
CRs-Fixed: 2915691
Add pld_get_mhi_state API to get current MHI state and
pld_is_pci_ep_awake to indicate that PCI EP is out of
low power state.
Also enable pld_prevent_l1 and pld_allow_l1 API for WCN6750
to manage low power state of PCI EP from Host.
Change-Id: Ic574bf7d7886b8e59b98e65806b387eef3a54e94
For reassociate case, when disconnect is triggered by
wlan_hdd_try_disconnect, then disconnect event is not
indicated to kernel in disconnect handler -
hdd_dis_connect_handler. And if the csr_roam_connect
get failure for some reason, the kernel will only get
error code from driver connect request interface and
the connected bss information in kernel will not get
cleared. Then kernel / driver will be out of sync.
To fix that, for such csr_roam_connect failure case,
let driver return successful status code from driver
connect interface and indicate connection result failure
event to kernel to clear kernel bss information.
Change-Id: I5367262751e63c04553deeb7607cbe13c70f8e15
CRs-Fixed: 2940994
Currently idle shutdown is flushed after the suspend dsc op
start. This can lead to the deadlock when the idle shutdown
work is running as it waits for the dsc ops to complete.
To mitigate this issue, flush the idle shutdown work before
starting the dsc ops in suspend sequence.
CRs-Fixed: 2941314
Change-Id: I023b9eb3543d83a30e0587cec3977e08ac7ab752
