Hif needs to guarantee that it frees the buffers that it doesn't
pass to upper layers.
Change-Id: I816d0a1b19a5a0c00e9669c18dfdf427c102c911
CRs-Fixed: 958124
This flag supports depricated emulation hardware.
Remove conditionally compiled code and dummy functions
that become unused as a result.
Change-Id: Ie21df7e39b196109acfb6b41afc6b8c22fecb822
CRs-Fixed: 958120
On M2M emulation platform we have a fixed mapping between macs, hence
vdev transition & MCC support is not possible on this platform. But MPR
platform doesn't have these limitations. This config allows at runtime
enable/disable vdev transition & MCC support depending on the platform
it is running on.
Change-Id: I17a92dce695ee30f7994f040d4bc612a38680f3e
CRs-fixed: 922181
qcacld-2.0 to qcacld-3.0 propagation
If ADDTS response timer is fired because of not receiving ADDTS RSP
within time then this timer handler posts SIR_LIM_ADDTS_RSP_TIMEOUT
command to LIM. But if deferring is enabled(say due to processing any
action frame) then SIR_LIM_ADDTS_RSP_TIMEOUT command will be added to
defer queue and as gLimAddtsSent is set to TRUE, including this
command all other defer commands will never be processed. This may
lead to active command timeout.
Hence process SIR_LIM_ADDTS_RSP_TIMEOUT command in defer state.
Change-Id: I48f09a7fd560702ac7701e0a1a41c77ad94626d4
CRs-Fixed: 919490
Fix compilation issues on MSM8998 RUMI platform caused by
CONFIG_CNSS not being defined and un-initialized local variable usage.
Change-Id: Iaf8bec493a42ec9fd09e1b2b6aa085be8de4bc8a
CRs-Fixed: 959249
qcacld-2.0 to qcacld-3.0 propagation
-In case CSR is already connected and it receives connect with the
same profile, csr will silently return and try to send the connect
success to HDD. But if roamInfo staId is 0 this roam command is not
removed from sme active list and leads to active list timeout.
This back to back connect to CSR can happen if HDD and CSR go out
of sync and is taken care already.
-Also if disconnect is received in case scan for ssid is in progress
it will be dropped and roam session state is set to disconnecting.
But this roam session state is not reset for the fresh connection
and thus if next connect also issue scan for ssid the connect
command will be dropped assuming disconnect is in progress.
This change removes roam cmd from sme active list in case of
above failure and resets roam session state for fresh connections.
Also does some cleanup in HDD disconnect command.
Change-Id: I222626f1da6bdad83a5264668dde9aafd3dda149
CRs-Fixed: 953196
qcacld-2.0 to qcacld-3.0 propagation
If connect is received with previous connect in progress, driver
doesn't abort the previous connect and process the new connect.
In csr if new connect is for same profile, sme directly indicate
connection success to hdd and thus hdd increment the active
session count twice for both connect.
Now if one more connect is received it will try to disconnect the
previous connection and decrements the active session count.
So the count is 1 though no session is active.
Eventually if same scenario happens again the active session
count will become 2 with no active session. As max active session
limit is 2, no more connect is accepted.
To avoid this, if fresh connect is received and previous connect is
in progress abort the previous connect.
CRs-Fixed: 958963
Change-Id: I1544bc6e7894ffeda1956a19a553b6fe4e35f988
qcacld-2.0 to qcacld-3.0 propagation.
If MC thread stall while waiting on any event, In case of WPS,
probe requests may get accumulated in PE message queue.
Now every time MC thread process a probe request,
eWNI_SME_MGMT_FRM_IND msg is posted in SME message queue. But
unless PE queue is fully processed SME queue won't be processed.
This will gradually use all vos message wrapper.
To avoid this register a callback to send management frame
indication directly to HDD from PE which avoids posting message to
SME queue.
Change-Id: Ib83700825112cc52dade594909bfa8993909ac29
CRs-Fixed: 944961
qcacld-2.0 to qcacld-3.0 propagation.
1.Use iw tool to let dut connect to remote ap, dut is working
on wlan0.
2.At same time, use hostapd to start sap also on interface
wlan0.
In item1, before connect to ap, dut will do an internal scan
to search ap, if scan result matched configured ap profile,
dut trys to connect to ap.
In item2, when starting sap, found it's sta mode it will try
to disconnect dut with remote ap and close sta session.
#1 If ap start action happened in item1's scan period, trying
disconnecting won't take effect as currently sta has not yet
connected with remote ap.
So after item1's scan finished, dut continues to do connection.
#2 During closing sta session, sme issued eSmeCommandDelStaSession,
as currently sta is doing connection, so session is in started
status, this cause closure failed, so there is no chance to remove
cmd in smeCmdActiveList.
While cmd pending in smeCmdActiveList for 120s, driver issued crash.
The solution is in item1 after scan complete to check whether it's
needed to start following connection. If stop action has been issued
during scan period, then don't do the connection.
Change-Id: If632e21cb13277fba371ff1bf8df97febf856d09
CRs-Fixed: 923375
Currently the HDD code uses a variety of logging APIs. In qcacld-3.0
HDD should converge on a unified set of logging APIs. Update
wlan_hdd_driver_ops.c to use the unified set of APIs.
Change-Id: I5620e91797412c0dbf32c1071b3eefd54f54aca7
CRs-Fixed: 937590
Currently the HDD code uses a variety of logging APIs. In qcacld-3.0
HDD should converge on a unified set of logging APIs. Update
wlan_hdd_lro.c to use the unified set of APIs.
Change-Id: I72bf2b55390e1c72706077d22f61e59d17d14f13
CRs-Fixed: 937662
Change "qcacld-3.0: Roaming enhancement to allow multiple ssid"
(Change-Id If4953e64f9e0485edec85167a7bc1dd6ca7c2047) added some
incorrect logging to function hdd_send_re_assoc_event(). The intent
was to perform a hex dump of the buf_ssid_ie and final_req_ie buffers,
but in reality the code actually uses these buffers as logging format
strings. Fix this by invoking the correct API to perform a hex dump.
Change-Id: Id861ececd119ea7247ce24c7c8ea9173f13703d3
CRs-Fixed: 958752
Currently the HDD code uses a variety of logging APIs. In qcacld-3.0
HDD should converge on a unified set of logging APIs. Update
wlan_hdd_nan.c to use the unified set of APIs.
Change-Id: I4dcc61656f1cd45d3a2c9d63441c27a733385328
CRs-Fixed: 937673
In adrastea multithreaded firmware WMI logging infrastructure
is not yet ready. Disable WMI FW logging for adrastea and enable
it for ROME based targets.
Change-Id: I64c858681f57a2a013fd23bcd9204f59ff638102
CRs-Fixed: 959416
qcacld-2.0 to qcacld-3.0 propagation
There is a race condition between TX_RESUME opcode, where checks
pending_cons_req before sending WLAN_CONS grant notification to IPA,
and WLAN_CLIENT_CONNECT event where clearing the pending_cons_req.
We don't need to clear pending_cons_req from WLAN_CLIENT_CONNECT
event, which is actually cleared by TX_RESUME opcode.
Change-Id: Icc5d05bc77800251b031efd0c07b0feaa2ae921d
CRs-Fixed: 952171
qcacld-2.0 to qcacld-3.0 propagation
When CDS enable fails after IPA init, IPA cleanup is missed.
This will cause a system crash later as IPA resource is not properly
teared down.
Change-Id: I563e92e4ed5319b6489dcc8cffb6a802d54e8549
CRs-fixed: 949015
Driver state of loading, unloading, logp are maintained in
multiple modules like HDD, CDS. Change to maintain the driver
state in CDS and provide CDS APIs to find out the state of the
driver so that any of the module can query it.
Also rename the logp to recovery in progress for clarity purpose.
Change-Id: I8e1864e1bc7f3b1dd6f4eb804ce2578c6695967d
CRs-fixed: 958659
FTM initialization has conditional compilation within the
function. Clean-up and create separate functions to initialize
and remove FTM during driver probe and remove.
Change-Id: I04890f49447c22b6d7dd93baf0a266ea9f5a4ca7
CRs-fixed: 957630
cdf_mem_exit cdf_mc_timer_exit are missing from hdd_driver_exit.
Add them back so that wlan driver exits properly.
Change-Id: I571f65118a1755d2f43c29e8320563a43b49e713
CRs-Fixed: 951519
Add event handler for oem data response event from FW.
Add new boolean member to differentiate the response
generated internally and one coming from the FW.
Change-Id: Ia906c5dcbb7e560fd78ccd96fc47c7801b33aa6a
CRs-Fixed: 942261
Use assoc req from roam_sync_ind_ptr instead of ft_session_ptr in
condition statement.
assoc req from ft_session_ptr is not populated yet and hence will be
0 when the condition is invoked. This will cause if block to never
execute.
Change-Id: I9cd758674d13603087de013a0aea751032bc8603
CRs-Fixed: 957591
Do not allocate txrx descriptors for management
if management over WMI is enabled.
Change-Id: Iafb231bebe83b658940119ed13a41d64a10d4eb0
CRs-Fixed: 955450
rssi_raw should not be updated when the channel
in probe response does not match the channel in BD.
Added rx_channel field to beacon descriptor passed from LIM to
CSR. Channel check is performed in csr_remove_dup_bss_description()
before computing new averaged rssi.
CRs-Fixed: 873549
Change-Id: Iefacbf73f880e9d0c957ad83839dd2203fdf862d
qcacld-2.0 to qcacld-3.0 propagation
Add raw rssi in the beacon header from WMI to beacon descriptor.
Pass it up to SME and report it to supplicant through a call to
wlan_hdd_cfg80211_inform_bss_frame().
It can be enabled as a .ini configuration parameter "gInformBssRssiRaw".
Default is 1, it will report raw rssi by default.
Roaming decisions are based on rssi field of beacon descriptor,
its interpretation and usage are unchanged.
CRs-Fixed: 815344
Change-Id: I6dacdc0b333c093d16a74f8cf36471dfc183ce56
If rxthread is on, NAPI will call the kernel-thread context API (instead
of softirq context) API to transfer the packet to the stack.
Eventually, this part will be re-considered when multi-queue NAPI gets
in, which will eliminate the rx-thread.
Change-Id: Ib3d4d1450fb36f5c1e710802ceaab5d015e20c7d
CRs-Fixed: 957299
qcacld-2.0 to qcacld-3.0 propagation
This is to fix the regression in the case of suspend request.
Below patch overkills the case for PNO and extscan in progress.
Commit Subject: "Fix the PNO WoW is not configured in concurrency case"
change id: I97084cfde73025f0d09b60616959c9530ba82c80
The combined condition check should be maintained to check
the case for PNO or extscan in progress, otherwise the WoW would
be enabled in the case PNO in progress is not enabled,
and end up not setting PNO patterns and failed to reconnect
after resume.
Change-Id: Icc5f19d07abea9fb25479452ce1c467d7811317c
CRs-fixed: 953257
This is for FR23278: Runtime configuration of softAP Tx power.
The value that send by the vendor command will set the power levels
as below:
0 --- no scaling, 100%
1 --- 50% of max power
2 --- 25% of max power
3 --- 12% of max power
4 --- minimum power
Another vendor command is to decrease the power by actual db value.
CRs-fixed: 941772
Change-Id: I61d48299b4af361c214e2ab8bb1fa848161431de
qcacld-2.0 to qcacld-3.0 propagation
Update common WMI commands to merge firmware init of MCL and
WIN SWOL update
Change-Id: I732a70be17d451654f3fbbe6db84f62de6438862
CRs-fixed: 865207
qcacld-2.0 to qcacld-3.0 propagation
Line 118: CID 50610 (#3 of 3): Out-of-bounds access (OVERRUN)
Line 200: CID 50680 (#1 of 1): Out-of-bounds read (OVERRUN)
Line 367: CID 50732 (#1 of 1): Out-of-bounds write (OVERRUN)
add TID validation check in RX_IND to avoid tid >19 scenario.
Change-Id: I32a3339902e231a8193fb8ee984ef029e24bc7e5
CRs-Fixed: 836067
Propagation from qcacld 2.0 to qcacld 3.0
Due to a race condition, vdev is getting deleted in
ol_txrx_vdev_detach and after that in
ol_txrx_peer_unref_delete, as vdev->delete.pending
is not equal to 0 as it is some garbage value, vdev
is getting deleted again causing crash.
This fix release the lock after checking
vdev->delete.pending so that before vdev gets deleted
in ol_txrx_vdev_detach, this check has been made and as
vdev->delete.pending is 0, it will not delete the
vdev in ol_txrx_peer_unref_delete and so vdev will be
deleted only once.
Change-Id: I4a7362b0e226c66ccb7c72243276d77e8df60b20
CRs-Fixed: 949397
Propagation from qcacld 2.0 to qcacld 3.0
The atomic variables are of integer type but "cdf_atomic_read",
"cdf_atomic_dec_and_test" and "cdf_atomic_inc_return" APIs
have return type of unsigned int. Because of this, these APIs
return positive value even when it is negative.
This fix change the return value of "cdf_atomic_read",
"cdf_atomic_dec_and_test" and "cdf_atomic_inc_return" APIs
from unsigned int to int so that negative value remains negative.
Change-Id: I9c4077547e1e1370bbb401af77b248944cdf25fa
CRs-Fixed: 902636
qcacld-2.0 to qcacld-3.0 propagation
Extended Supported Rates are filled in Assoc req based on the
dot11mode and in Case dot11mode is 11A it's not filled.
Some AP have Basic rate in Extended Rate set in 11A mode.
Thus AP reject the association as Driver does not sent the extended
rate set in 11A mode.
Removed the dot11mode dependency for Extended Supported Rates.
With this change if AP have Extended Supported Rates driver will
add the Extended Supported Rates IE in Assoc req.
Change-Id: Ibd1c98a7cfae46988ff4434579edde7d4222841d
CRs-Fixed: 903270
qcacld-2.0 to qcacld-3.0 propagation
If preauth is in progress with too many APs in the roamable AP
list and at the same time disconnect is received from supplicant as
part of unload, the deauth req will be queued in sme pending list
till csr completes the preauth process.
Now as part of unload, driver may purge the sme pending command list
and with it the disconnect command may also get removed if preauth
takes too much time.
This will lead to DEL BSS missing before HAL STOP and thus crash.
To avoid this abort the preauth process once disconnect is received.
Change-Id: I1c35f0c7e38295cfadafdad69f0b9a99ba61404c
CRs-Fixed: 942579