After every wifi up/down or load/unload, it is expected that
cfr enabled entries should be reset. Earlier, associated clients'
entires are cleared by FW. Unassociated entries are maintained by
host and were not cleared. This change clears them properly at
cfr_pdev_open level
Change-Id: I2dd7ce861901ccca72bc987c2f91677a481d401e
CRs-Fixed: 2531961
Target assert is observed due to sending insufficient MIC data for
GCMP cipher suite. So, Changing the MIC length for GCMP ciphher
suite from 8 to 16 bytes
Change-Id: I5a407f9eb0cca70b498d91eac2f60f1b8d16fb69
If radar is detected on the primary channel, agile DFS is restarted
after vdev start response by triggering a zero second timer.
The timer fires immediately and the existing agile precac channel
is marked as CAC done. Since the agile preCAC channel was never cleared,
it was marked as CAC done.
Reset the agile channel if radar is detected on the primary channel.
In case of radar in agile detector, the channel change will not involve
any zero second timers and the agile channel will be overwritten by a
new channel.
Also, consider the case where radar is detected on the primary
channel. HOST will now select a new primary channel and issue vdev start
to FW. After vdev response, HOST will pick a new agile channel.
Since the agile engine was never stopped, FW may issue OCAC complete
for the old agile channel at this instant. Now, since HOST has picked a
new agile channel, this new agile channel will be marked as CAC done
due to the OCAC complete event from FW, without preCAC being started.
Send OCAC ABORT command after radar detection to properly restart
the agile state machine and avoid any race conditions like above.
CRs-Fixed: 2528567
Change-Id: I455cd513ba612bf48f057583b7529c7c3ef2f6c6
When the current chainmask is updated, the FW ADFS support flags are
sent to DFS object in the API "dfs_set_fw_adfs_support". Update
the DFS flags representing the FW Agile DFS support based on the
new values.
Before starting (or) restarting agile DFS, check if user configuration
and FW Agile support is present and only then start the agile
state machine. If the current operating bandwidth of the pdev is
80P80 or 160MHz, check the FW flag "fw_adfs_support_160" and if
it's false, do not start the agile engine.
Change-Id: I0c44fd9d7d57042f82808979421284be02065c77
CRs-Fixed: 2521654
While checking if ETSI preCAC is already done on a given channel
in this API "dfs_is_etsi_precac_done", dfs_curchan is always used as
the channel to be checked. Since there are APIs such as
"dfs_is_cac_required", which may be called before dfs_curchan is
updated, add the channel structure which is to be checked for
ETSI preCAC in the argument list.
Change-Id: I77cf059f9afd83c850a7603f6197ac2b78fa2586
CRs-Fixed: 2528023
Signatures of target_if_dbr_buf_release and
target_if_dbr_cookie_lookup got updated as part of multi srng
support per pdev in DBR. Update the same in CFR.
CRs-Fixed: 2528137
Change-Id: If53fd9a1a4e4a30b8d9ba3eda9ade887e2c3322b
TLV and non-TLV definitions of vdev mgr ops are in component_dev,
they are carved out to a separate file in cmn_dev/wmi for adapting
target_if/vdev mgr component, they are made part of same file
which conflicts tlv design. This change moves redundant definitions
from component_dev to cmn_dev.
Change-Id: Ia47268b4479dba33dc2eceabeadf16224154a6f7
CRs-Fixed: 2517019
Consider the case where preCAC is running on channel 120 (20MHz)
bandwidth and 116 (20MHz) channel is already preCACed. One possible tree
structure, with the scenario mentioned above, is shown below:
122(1,0)
|
|------- 118(1,0)
| |
| |------- 116(1,0)
| |
| |------- 120(0,0)
|
|------- 126(0,0)
| |
| |------- 124(0,0)
| |
| |------- 128(0,0)
where 122 and 118 have 1 subchannel (116) as CAC done. While the preCAC
is running on 120(20Mhz), the preCAC is restarted with a new channel
of 40MHz bandwidth. If the new channel picked is 118(40Mhz) in which 116
is already CACed, once the preCAC is completed on 118 (40Mhz),
the entire band (with both 20MHz channels) will be marked as CAC done.
In the API "dfs_mark_tree_node_as_cac_done", the number of CACed
subchannel of the parent node is incremented without checking if the
child has been marked. In the above case, after preCAC is completed
on channel 118, each 20 MHz channels in channel 118 (116 and 120) are
marked individually in this API.
While marking the tree for preCAC completion on the subchannel 116,
the number of CACed subchannels is incremented to 2 for 122 and 118, but
124 is already marked. While marking the tree for preCAC completion on
the other subchannel (120), the CACed subchannels count is incremented
to 3 for 122 and since it cannot increment the subchannel count to more
than 2 for 118, it returns, leaving 128 unmarked. The preCAC tree
structure will now look like:
122(3,0)
|
|------- 118(2,0)
| |
| |------- 116(1,0)
| |
| |------- 120(0,0)
|
|------- 126(0,0)
| |
| |------- 124(0,0)
| |
| |------- 128(0,0)
which is wrong.
To fix the above problem, check if a tree node is already marked as
CAC done and only then increment their parent's CACed subchannels count.
This issue is not seen for other cases (e.g. marking NOL channels), since
preCAC can be performed on the same channel again (incase of different
bandwidths) whereas NOL marking is unique.
Change-Id: Ie683a011a86233dd2c1ff8e21aa78638a1d63c2a
CRs-Fixed: 2520790
If radar is found on primary channel, aDFS channel change is triggered
immediately after marking the primary channels as NOL. This may result
in picking the channel that the AP may come up in, for agile DFS.
To avoid running aDFS on the pdev's current channel, restart
agile DFS engine after new channel is selected, in vdev response.
Also, consider the scenario where the device has two DFS pdev objects.
When the API "dfs_prepare_agile_chan" is called to start preCAC, it
uses a round-robin logic to pick a DFS object for preCAC and
the preCAC timer maybe started for a different dfs object than the
one calling it (argument of the function). But the channel width of
the preCAC channel is always fetched from the dfs object calling
the API, which may result in wrong channel width sent in adfs_start
command.
Use temp_dfs structure (indicating the DFS object picked for preCAC)
to fetch the channel width, instead of fetching it from the dfs
object passed as argument to the function.
Change-Id: I243cf978b3caaadb771eb9a5e0cf83f7ac82c876
CRs-Fixed: 2515440
Grouping ppdu with same schedule command id into list
as there is no finite number on ppdu count with a same
schedule command id.
Change-Id: Ibf0fe7e29923b641ff6330c12200b351f42b6efb
While checking if the current channel overlaps with the
weather channel range, the first sub channel was checked if it is to
the left of weather channel start and the last sub channel was
checked if it is to the right of the weather channel end.
By the above logic, the API returns true only if the entire
weather channel range is within the current channel.
Fix the formula by checking if the first sub channel of the current
channel is to the left of the weather channel end and the last sub
channel is to the right of the weather channel start to find if any
of the current channel is a weather channel.
CRs-Fixed: 2515365
Change-Id: I8ca779f6dd8116a85121185e3ea71bca610f9ba4
While selecting a channel for aDFS, the current channel(s) are skipped
for preCAC (since they are CACed by the primary detector).
If no other channels are found and if the current channel
is the last in the preCAC list, aDFS is started on this channel.
This is because, the channel variable defined in the API locally
is not reset if it is an invalid channel.
Reset the agile channel at the end of the loop if it is one of the
excluded (current) channel.
Change-Id: I48cbfc3c511c05bed86b60eb9e2ca22f9b8e0362
CRs-Fixed: 2515324
As part of FR 56301, phase data is being captured in CFR.
Fill the extracted information in CFR header.
Change-Id: Ia1f6245a1a18d1d11b266328b17e6506bd8439f4
CRs-Fixed: 2501944
As part of FR 56301, add support to capture CFR data for
unassociated clients on its Probe response's ACK
Change-Id: I14a4eead4ceeb3eb21d8f2bdf76007f873db1f4e
CRs-Fixed: 2501938
Changing event dump size to 25 to make it same for all
events as qdf has limitation on event size.
If size is more than 25 then event of type command doesn't
gets printed.
Change-Id: Iaf2d4c52242b99194f7405d7d69a4566db71ff01
CRs-Fixed: 2510317
Smart Antenna support for extracting 11bit ratecode and packing
for tx packet training.
Change-Id: I8b838f3af8a6d2060b40c912e282dd3f11bf0339
CRs-Fixed: 2506480
Currently single API is used to get dfs_precac_enable and
dfs_agile_precac_enable. This API was wrongly returning value for
dfs_precac_enable since agile capability was set even for Cascade.
Sepate out the GET APIs for getting dfs_precac_enable and
dfs_agile_precac_enable so that getting these flags becomes independent
of each other.
Change-Id: Ie617c04269f91db32f851871880c4df884517a67
CRs-fixed: 2501266
Duplicate vlan tagged frames which need to be sent to repeater should
have unique check to avoid infinite iteration. Infinite iteration leads
to creation of nbuf every time loop is executed.
Change-Id: I2601461edc9e144f1058ed031ba14410563d0755
Move CONFIG files specific for AP into component_dev
instead of keeping them in cmn_dev
Change-Id: I8d44faec5b0e43d2197e1d08549637702be3848d
CRs-Fixed: 2506560
Currently, tx_monitor mode is enabled/disabled for all peers on a
per-pdev basis. Since tx_monitor mode consumes significant amount of
MIPS and reduces throughput, new requirement is to enable/disable
tx_monitor mode on a per-peer basis as well.
Change-Id: Ib70ccd3238dfd5fae1be92a81e9bf91fe441e228
CRs-Fixed: 2500931
Tags are programmed using wlanconfig commands. Rx IPv4/v6
TCP/UDP packets matching a 5-tuple are tagged using HawkeyeV2 hardware.
Tags are populated in the skb->cb in the REO/exception/monitor data
path and sent to upper stack
CRs-Fixed: 2470521
Change-Id: Id581159ae2b4a00c290a775aad526032e3a53a7f
While switching to 36 HT160 as part of multivdev restart, the channel
params to be sent to FW are only updated with primary segment dfs set flag.
Consider if the primary channel is 36 (non DFS), in which case, the
primary HT80 segment is non DFS, but secondary HT80 segment is DFS. This
will be indicated to FW by sending the flags with dfs_set as FALSE and
dfs_set_cfreq2 as TRUE. But since dfs_set_cfreq2 is not updated incase of
multivdev restart, FW does not detect radar in the secondary detector.
Add proper dfs_set_cfreq2 flag in multivdev_restart channel params.
Change-Id: I875534c3f61b1d9a36efe1ec63505332a59a4dd1
CRs-Fixed: 2496923
When CPU cores are fully loaded, PPDU stats process workqueue
is not getting enough time, resulting in excessive backlog of
PPDU status buffers. Check and drop buffers if the backlog is
beyond a threshold.
Change-Id: Ideef46cb4eec54dcd5a8315f36f2206653d81659
