During roam synch event processing, if ssid mismatch is seen,
then the driver issues disassociate, but still proceeds to
process the roam synch indication and indicates new
connection to upper layer. This causes policy manager api to
increment active session wrongly when actually STA is
disconnected.
Return failure in csr_roam_stop_network() if disassociation is
initiated.
Change-Id: I21ae1c26114e4387517710cd3358f89b7e2432fd
CRs-Fixed: 3210852
Currently shortgi 20/40 values are updated with ini configured
values irrespective of ap capabilities.
This change fills shortgi values based on ap capabilities if it is
enabled in ini.
Change-Id: I6a590e5c0981fab61dcbcf06cf22fda70d833fd7
CRs-Fixed: 2867232
Uninitialized variables in pkt_capture_process_tx_data()
and pkt_capture_callback() may lead to Unpredictable behaviour.
Initialize corresponding bssid variables to avoid it.
Change-Id: I1ca6e7e04c8920111414a83c29cb61883b72370c
CRs-Fixed: 3241885
qdf_freq_t is by default defined as uint32_t, but hdd_ctx->
unsafe_channel_list is defined as uin16_t because cnss interface
is defined as uint16_t, so can't pass hdd_ctx->unsafe_channel_list
to ucfg_policy_mgr_init_chan_avoidance() which expect qdf_freq_t
list.
Convert hdd_ctx->unsafe_channel_list to qdf_freq_t list and pass
it to ucfg_policy_mgr_init_chan_avoidance().
Change-Id: I0207827e3cbaa82526e97d2826d90d396148c84a
CRs-Fixed: 2964178
Sta is not allowed to connect/roam in 6 GHz frequency or indoor
frequency in non-DBS target if SAP is active.
But STA roams to 6 GHz AP when SAP is active since the PCL allows
6 GHz frequency.
While populating PCL to firmware, check if 6 GHz and indoor
frequencies are allowed for non-dbs target and set the
weight appropriately if the channels are not allowed
Change-Id: I0e5fdc5b3c4177283d91cdfc58359336cc11910d
CRs-Fixed: 3205494
In sch_gen_timing_advert_frame, the memory allocated for buffer
timing_advert->template_value is freed but not reset to NULL.
This creates a dangling pointer, and it is freed again inside
__wlan_hdd_cfg80211_ocb_start_timing_advert.
To avoid this issue, reset the pointer to buffer
timing_advert->template_value to NULL before returning from
sch_gen_timing_advert_frame.
Change-Id: I2445c53f217d0fd22cbe3026b0869284fe13b851
CRs-Fixed: 3229906
During SAE roam auth offload, update the beacon/probe response
frame coming from the FW via WMI_ROAM_FRAME_EVENTID into the
scan db.
When the bss info of the preauth candidate is unavailable in
the host/kernel scan entry, supplicant fails to determine the
proper SAE PWE config of the peer and the commit request fails.
Change-Id: Ia782ece7bebf8274f949fe9fb3b2a2882bf47597
CRs-Fixed: 3075458
The dynamic GRO enable/disable based on ingress filters
should currently is restricted only to standalone STA.
This was done to address issue where ingress filters are
added on STA interface whenever SAP is up disabling GRO
unnecessarily.
Fix is to allow dynamic GRO to work on all interfaces
and concurrency scenarios based on the priority of
the ingress filter configured to avoid unnecessary
GRO disable in STA+SAP.
Change-Id: I1742f4539353939e3a40ff4096b3f833f2029b12
CRs-Fixed: 3206815
In case of STA+SAP, ingress qdisc filters are getting
configured on both STA and SAP interface which is
causing GRO to be disabled.
Check for ingress qdisc to enable/disable GRO only in
case of stanalone STA and not do this in case of
concurrency.
Change-Id: I3542930c7b14d72e267378dd4687ee9721eed4ed
CRs-Fixed: 3062195
Currently unprotected diassociation frame is dropped if
PMF is enabled and keys are not installed. This causes
delay as supplicant has to do EAPOL timeout after
10 secs and then do disconnect.
Allow unprotected disassociation frame when install keys are
not done.
Change-Id: I12068d7b10c4e6f7885ca3a3400a8845a4c236f2
CRs-Fixed: 3208427
When PMF enabled, SA Query will be triggered unexpectly if
duplicated assoc_req received after pre_auth node deleted
and before key installed.
Here drop such duplicated assoc_req frame.
Change-Id: I1c769892052fba5e8c7340b2f82e822add573dd4
CRs-Fixed: 2966447
Currently duplicate authentication frames are dropped if the
sequence number is same previously processed auth frame and retry
bit is set. To stop the flooding of auth frames with same sequence
number, check the previous seq number stored on the preauth node
and drop if the frame is a duplicate.
Currently, for PMF-capable peers the duplicate assoc request
frames are getting processed before the key is plumbed because
the is_key_installed flag is not reset properly before assigning
the dp hash node to a new sta. Therefore, reset the is_key_installed
flag for every new dp hash node addition.
Change-Id: Id035a5ba5ffe9b745361956d05496f9905bf2526
CRs-Fixed: 3155517
In new framework the 6 GHz preferred SAP ACS config file will
contain 5 GHz channels, to avoid 6 GHz non-PSC being set to
MAX weight value and 5 GHz channels selected easily, here set
normalize_acs_weight to 90 for non-PSC channel of 6 GHz, 40
for 2 GHz and 5 GHz channels.
Change-Id: I5e89339d0fd4e3c190dd2ead1ec87d12ff197691
CRs-Fixed: 3186587
This change enables the CONFIG_ENABLE_LOW_POWER_MODE
based on the new flag added for the talos target.
This Low power config enabled the Deep Sleep / Hibernate
feature in WLAN.
Change-Id: Ied9a52d3b225957d0c3121df50bcd43d0ce7c7fc
CRs-Fixed: 3219080
Vendor commands of wiphy aren't initialized when driver load as FTM mode,
if changed to mission mode later with following cli:
echo 0 > /sys/module/wlan/parameters/con_mode,
vendor commands of wiphy keep uninitialized, wpa supplicant can't get
roam offload status by QCA_NL80211_VENDOR_SUBCMD_GET_SUPPORTED_FEATURES,
so PMK isn't sent to driver/firmware, roam offload failed.
To fix it, remove QDF_GLOBAL_FTM_MODE check in wlan_hdd_cfg80211_init, let
vendor commands of wiphy initialized in FTM mode too when driver load.
But in each vendor cmd, check FTM mode when called, if FTM mode, return
-EPERM.
When driver is changed to mission mode, all interfaces are deleted and
re-added by driver, when interface added event is received by
wpa supplicant, QCA_NL80211_VENDOR_SUBCMD_GET_SUPPORTED_FEATURES is
called again to get roam offload enabled status, PMK will be sent to
driver and firmware when RSO start, roam offload will work.
Change-Id: Ie0c868ebd4b7e356d72a154206b13df94f378188
CRs-Fixed: 3213221
To use skb_orphan instead of skb_unshare, this is
aimed to prevent addition skb_alloc possible failures
in TX path, then avoid of unnecessary packet drop when
memory runs low.
Change-Id: Ic8dfdb09c73a1071678612430fff2f23180ad336
CRs-Fixed: 3162137
NDP peer gets created as part of NDP indication in case of NDP
initiator. But NDP may fail to form due to various reasons and
firmware may send NDP confirm with reject status in such cases
instead of NDP indication event.
Below is the detailed scenario,
1. On initiator side, it has sent an NDP request and is waiting
for NDP response.
2. On responder side, application/framework is preparing to send
NDP response with reject due to some config mismatch. At the
same time, NDP termination is also issued. So NDP termination
frame is sent successfully.
3. Initiator firmware receives this NDP end and it sends an NDP
confirm with REJECT to the host and it shall clean-up the
session.
Currrently, there is a check for the peer existence while
indicating this NDP confirm status to framework. There is no need
of this check in such failure cases as peer is not yet created.
Change-Id: Ibe94a5b67df1ce3d65eaf2ef37b11b08155752c9
CRs-Fixed: 3086975
In api hdd_update_nss(), when unit test command 'iwpriv wlan0 nss 1'
is used in 1x1 mode, the fail status is returned resulting in test
case failure.
Return a success status in api hdd_update_nss() for 1x1 mode resulting
successful execution of test case.
Change-Id: Id482c2686b48a0b6cdd41980c4eb4424da775ad0
CRs-Fixed: 3195517
For session_id is not have sanity check, an OOB would happen
leads to crash.
Added Sanity checks for session_id in roam functions and
return failure if session_id is equals to NULL or MAX value.
Change-Id: I033c6567f53ead2a721882e9b44f3d0bcb58084d
CRs-Fixed: 3149880
Notif_param of Roam event ID carries the trigger reason as part
of roam scan start notif. Currently host doesn't read this param.
This commit reads the trigger reason and sends it to user-space
via vendor attribute whenever the roam event vendor command is
enabled.
Change-Id: If9fdbc31df149f4f1c9ca196ee74e2fcb3709a1c
CRs-Fixed: 3161803
Currently, wakelock is only taken during initial connection if
pairwise key installation is pending and not during roaming. This
may lead to EAPOL delay/timeout during roaming as runtime pm and
suspend is allowed during roaming. To fix this issue, acquire
wakelock if key installation is pending during roaming.
This change also refactors the code and release the wakelock in
wma_remove_peer if EAPOL fails during initial connection or roaming.
Change-Id: Id4cac30b3c383ca3d3e963571846f8a30eaa1006
CRs-Fixed: 3189799
Firmware might send NDP peer indication/NDP responder response
while NDI deletion is in progress. Currently, NDP peer entry
gets added though vdev is moved to DOWN state and cleanup is
in progress. This leads to stale peer entry on that NDI vdev
and doesn't allow vdev delete when IFF_DOWN comes, as a peer
is present.
Check the vdev state before creating peer for NDP and skip
peer creation if NDI vdev is not in UP state(which indicates
that the vdev down is happening).
Also, don't process the NDP confirm if NDP peer is not created.
Change-Id: I73e79cc33b1a294e3f4a4a6e2f371033862c5709
CRs-Fixed: 3061756
With Kernel 5.13.2 onwards, if
ppe_present is present in HE capabilities
then ppe_threshold fields are also required
to be sent such as length has to be non zero
and the data can be all zeros as part of
TDLS Discovery and Setup frame.
While sending setup frames, ies are packed using
dor11f apis such as dot11f_pack_ie_he_cap().
Even if ppet_present is set,
dot11f_pack_ie_he_cap() will only send
ppe_threshold if num_ppe_th is present.
The HE capabilties which are sent in these
are filled using MLME Cfg HE capabilities.
So, calculate and set num_ppe_th using cfg
ppe fields based on the operating band.
Also to cover a corner scenario where ppe_present
is set but calculated num_ppe_th is 0, set
num_ppe_th to 1 so that atleast 1 byte should be
sent to prevent framework from throwing
exception during add_sta.
Change-Id: Ibe4fd3fdacce24573c09b3609babf04ef4423fb1
CRs-Fixed: 3183531
Currently STA can scan and come up on 6 GHz or 5 GHz indoor channel
if hardware is non-dbs and SAP is present
As part of this change, do not allow STA to connect on 6 GHz and
5 GHz indoor channel for non-dbs hardware if SAP is present
Change-Id: I48ed8483225e35d8b898b0325eee398f74e41c97
CRs-Fixed: 3196247
While processing roam synch event, host driver reads the
reassoc request frame IEs after the SSID offset. SSID offset
is different for reassociation and association frames.
Currently SSID offset is populated as association frame ssid
offset only if ft_im_roaming flag is set. But FW can send
association request in roam invoke to self cases also and
host reads the ssid offset as reassoc frame offset since
ft_im_roaming flag wont be set here.
This causes association frames passed to upper layers be
truncated.
Read the assoc frame IE based on the frame subtype.
Change-Id: If927e294e9809b6640e4de23d97a7637b16db70c
CRs-Fixed: 3195902
Read early raom scan stop threshold min and max values
correctly and send correct value to FW.
Change-Id: I2e96b3c3e9fb48352d8075a4e7a96eae41dacedc
CRs-Fixed: 3191076
This change updates the auth_status and mac address of roaming
peer received in roam_sync_indication from FW to dp.
Change-Id: I27744017c87a956927108c5b3b0f9220d7e42965
CRs-Fixed: 3185110
Currently, LIM_ACK_NOT_RCD is not handled while filling
status code due to which STATUS_UNSPECIFIED_FAILURE is
updated as part of default handling instead of
STATUS_AUTH_NO_ACK_RECEIVED. Don't overwrite previous
status code in case no ack received from peer.
Fix is to Handle LIM_ACK_NOT_RCD case while filling status
code and consider previous status code when no ack received
from peer.
Change-Id: I366d63ee3ff6abf8a36300451e7f0c46fcf47e39
CRs-Fixed: 3151992
