Current HAL delayed reg write is tied to SRNG notions, hence
implement delayed reg write logic in HIF since WCN6450 does
not use SRNG interface.
New feature flag FEATURE_HIF_DELAYED_REG_WRITE is introduced
to disable/enable this support.
Change-Id: Id7087ad53cd5879cf49ee0e84dd727de61137541
CRs-Fixed: 3519702
Define a new QCA vendor attribute to enable/disable trigger
based UL MU transmission.
Change-Id: Ide36ec4869dc462879a57ff9a1e06f9f48e5c8d0
CRs-Fixed: 3538865
Regarding to the platform suspend, generally the hif_bus_suspend() will
disable the CE irqs, however the irq is shared in the single MSI case,
so the MHI irq is disabled as well which cause no MHI ring event notified
in the following pci bus suspend.
The fix is to bypass hif apps irqs handle for single MSI case during
suspend-resume stage.
Change-Id: I4f9068900e85fe57c2a82d841846b3fff6da0696
CRs-Fixed: 3549884
Masking the MLO API, mlo_setup_get_total_socs
using ic_eht_target check for non-MLDs in cmn_dev.
CRs-Fixed: 3551482
Change-Id: I8d585f534146f58beee605b2165bb0c51fc5bdb9
Currently policy mgr api policy_mgr_convert_device_mode_to_qdf_type
accepts only QDF_OPMODE as an argument and converts it to
policy_mgr_con_mode. Also, name of this api does not reflect
the task this is doing.
With the new SAP mode for low latency low throughput SAP,
to convert the QDF_OPMODE to policy_mgr_con_mode, host driver
needs to know if it is ll_lt_sap or normal sap. To get this
information, update policy_mgr_convert_device_mode_to_qdf_type
api to accept the psoc and vdev id as parameters. Also, rename
this api to reflect the task this apiis doing.
Change-Id: Ib182d8091305fd9c4f35b424d7b4a751a61730ad
CRs-Fixed: 3525427
When one of the enabled CE tasklet is executed, it will
check latency of all the others which are enabled, it's
redundant. Add a new API to check latency for a single
tasklet.
Add API to record sched time for a tasklet, which will
check the enablement of detection before recording the
time.
Change-Id: I515ca27fc2e97231397573b4903c001d29326bd2
CRs-Fixed: 3514285
When the peer delete timer/vdev manager response timer expires, host checks
if the umac reset is in progress. If so, host will retrigger the timer.
It is possible that the umac reset is just completed and either of these
two timers got expired. In this case, host will assert saying that the peer
delete response/vdev manager response is not received from FW. FW may take
a few milliseconds to send the peer delete/vdev manager response to the
host after the umac recovery completes.
Hence, to avoid the crash, check if the umac reset was in progress during
the umac reset buffer window.
Add the below changes as well,
1) INI support to get the umac reset buffer window value.
2) Renamed the API dp_umac_reset_is_inprogress as
dp_get_umac_reset_in_progress_state to return either
CDP_UMAC_RESET_IN_PROGRESS or
CDP_UMAC_RESET_IN_PROGRESS_DURING_BUFFER_WINDOW.
Change-Id: Ie15ef0731bad4b0ed955a479f00e297b7ba10729
CRs-Fixed: 3522665
This change initializes some uninitialized local
variable to avoid arbitrary values in some cases.
Change-Id: I906528514c41132ead23433a8a91374b4309c543
CRs-Fixed: 3545869
The change 'I4acd2f70707386f731e9b8fa2ef90a37868cbe23'
makes modifications in MLME and DP (Data Path) to optimize
the check for radar in per packet transmission path to
reduce CPU processing overheads. This new check is based on
a VAP variable that indicates whether radar is detected
or not.
Upon radar detection, send an indication to MLME through the
registered callbacks to update the status of the flag.
Change-Id: Id0582b5d79be672949f150907eb9ee64fa17d4f7
CRs-Fixed: 3528956
Currently RTPM refcount is not released if force
wake release fails. This leads to mismatch in number of
RTPM get and put during RTPM de-init.
Fix this refcount release if force wake release fails.
Change-Id: I4334de57e00139c0d30d37d7e19a399760400a50
CRs-Fixed: 3544981
Add missing nbuf_count updates in qdf APIs which are used
for keeping track of nbufs in debug mode.
Change-Id: If031ec8860513dddb1704a75e4761ce14aa26f21
CRs-Fixed: 4728458
Remove alignment of 4 bytes when allocating memory for Copy
engines. Copy engine buffer size is 2048 which is already
aligned to 4 bytes. This 4 byte alignment addition results in
64 bytes extra memory allocation for each skb since SKB_DATA_ALIGN
rounds it off to nearest 64.
Change-Id: Ib680547d2bcfe57b6eadda5301677d68e4a931a4
CRs-Fixed: 3533492
All packets except TID 0 first come to REO error NULL queue.
This happens because only TID 0 queue is setup initially.
Added multipass specific code in NULL queue path.
Change-Id: I8103215ad3e28796c868caa3eec816b9d118da9d
CRs-Fixed: 3482796
Add API utils_dfs_get_radar_bitmap_from_nolie() to fetch the puncture
bitmap from NOL IE. Also, move dfs_convert_wlan_phymode_to_chwidth()
from oss to common code. Also, add a 11BE version of
utils_dfs_convert_wlan_phymode_to_chwidth() API.
CRs-Fixed: 3538818
Change-Id: I258452e122d259c7b16655b760f217073b8e2e9b
Change number of packets datatype to unsigned 64bit and handle
respective print message changes.
Change-Id: I65e1ae90985b7d2fb29ea57f8034f5858491dae1
CRs-Fixed: 3534513
Currently, IPA SMMU map/unmap is called as part
of init. This causes every nbuf to be mapped
to IPA in the Rx path, causing throughputs
to drop. This change resolves the problem by
deferring the IPA SMMU map/unmap
call to OPT_DP filter reserve/release, as
nbuf needs to be mapped to IPA only in this scenario.
Change-Id: If198a6c5f22af58fdaf9d9c020c74b1f76002e37
CRs-Fixed: 3496679
Trigger the STA primary umac update event to umac
to handle downgrade scenarios.
Change-Id: I510d3d4443a978a8e2112f7372d0d1a03a508d18
CRs-Fixed: 3540314
RootCause:
WBM RX Error Ring, receives a packet from RXDMA with FCS Error
and MSDU Continuous Bit set. As MSDU continuous bit is set the
nbuf is added to temp SG list, but SG for these error is not
supported, hence the temp SG List never got reset and it
incremented hugely leading to high number of nbuf for
processing in list which lead to RCU stall
Fix:
IN WBM RX Error Ring, SG is supported only for two kind of
errors i.e. NULL Desc and Unencrypted. For all other kind of
errors reported by HW with MSDU Continuous bit set, explicitly
reset MSDU continuous bit in nbuf
Change-Id: I8d5858ccefb89b93e1c4fb36609d747eef00e8f2
CRs-Fixed: 3532208
Add support for per-MLO link configurations in
QCA_NL80211_VENDOR_SUBCMD_SET_WIFI_CONFIGURATION and
QCA_NL80211_VENDOR_SUBCMD_GET_WIFI_CONFIGURATION commands.
Additionally, add documentation for
QCA_NL80211_VENDOR_SUBCMD_SET_WIFI_CONFIGURATION and
QCA_NL80211_VENDOR_SUBCMD_GET_WIFI_CONFIGURATION commands.
Change-Id: Id83912079b3a4876891ac48beb069b17d4b8ebbf
CRs-Fixed: 3540412
Add support to control maximum allowed bandwidth update type for
the current connection.
Change-Id: I27880a5a969adf46861a58907d1e9e8ec35f9053
CRs-Fixed: 3540098
ipa_ctx->sys_pipe[] is an array with size WLAN_IPA_MAX_SYSBAM_PIPE (4).
On MDM platforms, WLAN_IPA_MAX_IFACE could be set to 4 in defconfig.
E.g. on sdxpinn/hmt platform, CONFIG_NUM_IPA_IFACE is set to 4.
kiwi_v2_defconfig
CONFIG_NUM_IPA_IFACE := 4
With this, WLAN_IPA_RX_PIPE will be equal to 4, leading to array index
out of bounds buffer overflow.
Fix is to set WLAN_IPA_RX_PIPE to WLAN_IPA_MAX_SYSBAM_PIPE - 1 to point
to last index of ipa_ctx->sys_pipe[].
Change-Id: I4929aec92ed2064c5194ec18ad7b33574ba3c3e7
CRs-Fixed: 3545420
Currently in host support is there for SET_FCC_CHANNEL
command for values 0, -1, 1.
Add changes to support SET_FCC_CHANNEL command for value 2.
Change-Id: I151ce33677406670bebc7a8861e34da3e4c5917a
CRs-Fixed: 3537395
On sdxpinn target (first multi-core MDM platform) with rome
chip attached, it's required to enable IPA pipe for both SAP in
AP-AP mode, but currently only first AP has IPA pipe enabled.
The second AP always goes with sw path because wlan firmware isn't
notified to choose IPA path by wlan_ipa_uc_offload_enable_disable().
The fix is to always enable IPA for each SAP no matter how many
SAP are started if MDM_PLATFORM is defined.
Change-Id: I8a2048aae04e14aa840d7d9a2ccb287201336cf3
CRs-Fixed: 3532926
Fix the peer authentication for wds client, in SDX_WKK
wds client needs to get authenticate when packets receive
from exception path as wds client is not directly associated,
authentication is done by verifying from ast table
Change-Id: I93051d67fc3d5be0af9242e6579f44883b3f757d
CRs-Fixed: 3534370
Host driver is currently indicating link reconfig
event on link vdev.
Change is to indicate link reconfig on the primary vdev
affiliated with wlan0.
Change-Id: I9955d98bcd52554adf22f838bfbd2263418432d1
CRs-Fixed: 3535940
For LPC mode, host only need to configure the filter for
monitor status ring, filter for STA RXDMA_BUF has been
configured during dp_soc init, do not touch filter of
RXDMA_BUF for LPC mode.
Change-Id: Ia64b0f608ce7c31517ef7174293bc52e6d4ab9b7
CRs-Fixed: 3531963
For 11A only AP, HT and DS IE are not present in beacon,
STA may receive the beacon on adjacent channel, then
driver may record wrong channel frequency in scan entry.
Fix by update the channel frequency by old scan entry
channel frequency if new beacon RSSI lower than old beacon
RSSI and gap is bigger than 40 db within 5s.
Change-Id: Idbd144befa78597fcd50a753ca13b0f11d1f45f5
CRs-Fixed: 3484664
Add fix to set NULL to msdu extn descriptor on release.
if same descriptor is used for other frame type. there is a more chance
extension descriptor can be added to the same free list.
Change-Id: I25e86c0ae03b4b39b6813519864afd16b3d19e4c
CRs-Fixed: 3503237
In case MLO HW will do the link mac address to mld mac address
conversion in RX path, But currently HW is supporting only for
802.3 mode.
For RAW mode(802.11) this conversion not happening on HW so added
this address conversion logic on SW.
Change-Id: Ibd794510e9d0ebdca884ed2e4043e0c48052156e
CRs-Fixed: 3537914
When AP is moving to down state, stop the puncture SM.
If we do not stop the SM, the corresponding puncturing
slots will not be available for new radar puncturing.
Introduce a new member dfs_is_unpunctured in dfs_punc_obj
structure. The dfs_is_unpunctured holds the information
whether SM object is unpunctured at that moment. Since
both un-puncturing and puncturing are done through vdev
channel change, current code cannot differentiate between
if the channel change is for puncturing or it is for
un-puncturing.This variable will help us identify whether
an event should be for puncturing, or it is for
un-puncturing.
Handle the radar puncturing after the usenol check,
otherwise the usenol case will not get executed.
The internal_radar_pattern is not needed as the
information can be fetched from DFS puncture SM object.
CRs-Fixed: 3540171
Change-Id: If889348056ad82e7cbbe6ca954af7ddfa99e6865
Add send_mlo_link_set_active_id_cmd_tlv API to send
force link bitmap to target.
Change-Id: Ie3a41bd2c82b6d10c8d82e4afa83621969c01798
CRs-Fixed: 3515441
Swift radio in Hawkeye is a partial offload chipset supporting
11AC mode. It has no support for MLO. Currently, when MLO event
handler registration fails due to No Support, an error is returned.
This terminates the device bootup causing a crash.
To solve this issue, do not return an error status when the
event registration fails due to No support.
Change-Id: I89819461bbc1d0fa31ef0558c93d06723f13de35
CRs-Fixed: 3520749
If AP doesn't advertise 6 GHz power type or advertises
invalid 6 GHz power type, find the power type for connection
as per below logic.
a) Connect in VLP only if VLP power type is available for STA
for the connection channel.
b) Otherwise connect in LPI only if LPI power type is available
for STA for the connection channel for STA.
Don't connect if VLP or LPI power types are not available for
STA for the connection channel.
Change-Id: I4c2d146050bc49b856f22c15af36a2b1d229f448
CRs-Fixed: 3529927
