Extract agc gain information, CFO & rx_start_ts from
rx status and cfr peer capture event to CFR host metadata
Change-Id: Id896554f1da6920bd3eeced1298b07857d41a68b
When the option -Werror=tautological-compare is enabled, the compiler
reports error because bitwise comparison always evaluates to false.
Change-Id: I7532dc9bccb69e02be5f04897a6ecaff44e84a0d
CRs-Fixed: 2734883
The 'timestamp' field in CFR meta data corresponds to TSF
at the reception of PPDU in micro second unit. To capture
timestamp for one shot, periodic and repetitive channel
capture, this field is being used.
case 1:
For one shot and periodic capture, the clock used for
timestamp is private to the target and not visible to host.
So it is being passed by FW to host and the placeholder
being exchanged between Host and target is of 32 bit.
case 2:
In case of repetitive channel capture, the
timestamp is being populated from the received PPDU,
which is 64bit.
To accommodate the timestamp, in both above cases, change
the data type to u_int64_t from u_int32_t
Change-Id: Ifa612bd97dc4f597368daefcd4224b84f2484c03
CRs-Fixed: 2751408
Enable CFR support for IPQ5018.
Calculate stream FS buffer length based on CFR payload,
CFR uCode header and CSI metadata.
Change-Id: Idd903ead75d635ce87ffc0df80e5c32fb62d15aa
CRs-Fixed: 2751408
Repetitive channel capture configurations usually support
16 groups to be confugured. To check if a particular group
is enabled by user, qdf_test_bit is being used here.
qdf_test_bit usually operates with uint32_t and it takes
the argument as unsigned long type. Since,
modified_in_curr_session is declared as uint16_t, and then
typecasted into unsigned long.
Change the data type of modified_in_curr_session
from uint16_t to 'unsigned long' to avoid compiler
issues.
Change-Id: Iec778dc36d5671f2cfb643c9ced1bd2a6104de2b
CRs-Fixed: 2753267
Update the signal strength calculation logic in CFR to consider
negative SNR i.e., a signal below the noise floor.
CRs-Fixed: 2753751
Change-Id: I2e2b95cc2ac2cf50ad5b7e38977ab971ddeb1ecf
Add freeze_tlv_version in cfir_enh_dma_hdr where,
version 1: Cypress/HastingsPrime which supports single user
channel capture
version 2: Maple/Spruce/Moselle which supports MU channel
capture for up to 4 users
version 3: Pine which supports MU channel capture for up to
37 users
CRs-Fixed: 2741027
Change-Id: Iadd6dc2ff82030f502776ce35988b4472450337e
Support a new capture_interval_mode: capture_count, where after
capture_count+1 number of frames, MAC stops channel capture and
waits for capture_interval duration before enabling again.
Add nob(capture_intervalmode_sel) to switch to capture_count or
capture_duration mode in wlanconfig application, where
capture_intervalmode_sel=0 indicates the old capture_durarion
mode where MAC stops channel capture after capture_duration
instead of capture_count. capture_intervalmode_sel=1 indicates
the capture_count mode.
Support MO marking: Current ENH CFR uses the existing MD/MO,
type/subtype filters to filterin the packets with programmed
type/subtype as either MD/MO. But for m_TARA it is also required
to consider the bw/nss/ta/ra setting. Therefore, it is possible
that not all the filtered-in PPDUs have channel capture and lead
to processing of packets which does not have captured data.
To overcome this issue, from QCN9000 onwards, MAC has MO marking
feature added for M_TARA filter mode, where MAC will consider
all TARA group configuration as filterin criteria & provides
an option to HOST to choose either filterin mode as FP or MO.
en_ta_ra_filter_in_as_fp is added to support MO marking.
Based on the discussion between HOST, FW & MAC teams, 0 is the
recommended & default setting for en_ta_ra_filter_in_as_fp.
Add nob to configure en_ta_ra_filter_in_as_fp but do not expose
to user.
Change-Id: I790d7ce500a7d779ea571635690f73dbc582cf33
For adrastea targets, enhanced CFR is not supported. When only
CONFIG_WLAN_CFR_ENABLE is set, some compilation errors are
observed as CONFIG_WLAN_ENH_CFR_ENABLE is not set true.
Fix these compilation errors when only CONFIG_WLAN_CFR_ENABLE
is set for targets which do not support enhanced CFR.
Change-Id: I96a769d8f10144033e48e123bb68d48a54cdf8ab
CRs-Fixed: 2706904
Decouple legacy FW headers for lithium builds since
its not required for the same
CRs-Fixed: 2719782
Change-Id: Ibdfdec51c5379fe42e23bf04a228ae6fb1a3eac9
Add the init and deinit CFR pdev api for adrastea target type.
cfr_adrastea_init_pdev api enables the CFR feature if both host and
firmware supports the feature, initialises the cfr pdev structure
and streamfs.
Also during allocation of memory requested by firmware, skip smaller
chunks memory allocation for CFR capture requested memory as firmware
expects contiguous memory.
Change-Id: Ib893dc8a845fea990a11e95e98d8a656f7df12ba
CRs-Fixed: 2687057
CFR component allocate 140 luts but some of platform use 48 of them,
which allocate huge memory at one time and waste memory in some
platform. In low memory case, it may be failed to alloc luts at one
time. So, allocate lut based on number of DBR entry and split memory
allocation.
Change-Id: Id49c3f9a0f81cfa6d639cb72b62bb377911f0e77
CRs-Fixed: 2679470
In some scenarios, where cfr data is being captured
for maximum supported bandwidth case, memory allocated
in case of streamfs for all 255 buffers is not enough.
Corrections are made for maximum capture length.
Maximum capture length for a single CFR capture should be
(max bandwidth case) 16240B (112 bytes(csi header)
+ 64 bytes(cfr header) + 16064 bytes(cfr payload))
Remove debug logs under spin lock
Change-Id: I7cdc25350f3d85ecb7d0d69431c3c66dbc9b2bed
CRs-Fixed: 2678750
It uses point to 16 bits variable as 32 bits pointer when setting ta
ra configurations to default. This causes memory access out of
bounds.
Change-Id: Ifa7adb0249cb23fdf71e53f04b92b62c1b215632
CRs-Fixed: 2677422
When user disables CFR through INI -
1. Skip CFR SRNG allocations
2. Skip WMI handler registrations for DBR/TX completion events
Change-Id: I4f397e35717dc492aea5af74b167c1111cbc16a9
CRs-Fixed: 2671611
The number of rx chains in target is different to the value in host
side, this change converts the value from target to host.
Change-Id: I86044bf12e958da312924827a3fd1e6799beaf41
CRs-Fixed: 2668403
FW requires to receive WMI_CFR_CAPTURE_FILTER_CMDID with the pdev id
which is based on the working band. It requires srng id when looking
up or releasing DMA address. So add this change to check working band
and set mac id. This is MCL only change.
Change-Id: I763a25b4989607128b2b4c75186eb5bdcd204077
CRs-Fixed: 2641268
CFR uses the number of chainmask and bandwidth from DMA and updates to
userspace.
Change-Id: I8f54cdbfdf77d39f9862848d3cacb61eb4924a05
CRs-Fixed: 2644015
Add chip type for QCA6490, and CFR component will indicate chip type
to user space.
Change-Id: Ia3e95e06c8cc86bc30b81906f62f3821606a38f7
CRs-Fixed: 2659553
1. MAC does not really consider the expect TA and the
expect RA when the expect TA mask and expect RA mask is 0.
Basically, it checks for the subtype, bandwidth and NSS
as the filter criteria.
2. In the other scenario, if the RA matches with expect
RA and TA matches with expect TA then it does not really
consider the expect TA MASK and expect RA MASK.
With the current default config, since the expect address
mask is 00:00:00:00:00:00 it falls under case 1 and it is
written based on the MAC recipe.
But actual MAC logic is different from what is mentioned
in the recipe. MAC works on below logic:
TA & TA MASK = Expect TA & expect TA mask
RA & RA MASK = Expect RA & expect RA mask
As an additional precaution, we need to modify default
subtype as 0 so that if user enables any group by mistake
and does not give any input for the other RCC
configurations, we can block the RCC to capture CFR data.
Change-Id: I759e8812e4ef0e77f773d0d445134e62d9e8a753
CRs-Fixed: 2649714
Protect access to cfr lookup table, since error handling
mechanism attempts to cleanup CFR entries in timer context,
this leads to contention between bottom-half and timer.
Change-Id: I32a095fc82545ab117ab43a81ef8a3a5602741fc
CRs-Fixed: 2649714
1. Clear cfr_timer_enable flag during wifi down
2. Similar to one-shot capture, report RSSI in metadata for RCC.
Convert SNR in rxtlv to RSSI.
Change-Id: I1802ea25b306faf06be52da60edb5050ff1bea9c
CRs-Fixed: 2649714
TA retrieved from freeze TLV was being over-written by macaddr
from AST search during RXTLV processing.
MACADDR from AST search is essential only for MU PPDUs.
Change-Id: I1a7c46dd11b6cae6049ae70046af849228ba3b47
CRs-Fixed: 2649714
Skip the freeze data and capture type related processing
if freeze data is invalid.
CRs-Fixed: 2637644
Change-Id: I97da3b9dc0d655e47dec534d04e0897af57e33d8
This is part of moving CFR component to hostcmn. Fix compilation
issues in this change.
Change-Id: Ia0a40da1e7e9201ffc5d1700cccea4181b358842
CRs-Fixed: 2637129
Since there are chip differences about channel frequency response. Add
files for QCA6490. This change also adds CFR configure file
Change-Id: Ib200affe4c68dac5535c4a3d65109072559161ff
CRs-Fixed: 2637134
