| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
| |
PLT firmwares *.3.10.0.142 for single-role
were corrupted and not usable.
Upload new PLT firmwares instead.
Signed-off-by: Guy Mishol <guym@ti.com>
|
| |
|
|
|
|
|
|
|
| |
Added WPA2 Krack fix for both SR and MR FWs.
SR: X.3.10.0.142
MR: X.5.7.0.50
Signed-off-by: Guy Mishol <guym@ti.com>
|
| | |
|
| |
|
|
|
|
| |
If we got a beacon with a longer TIM we overwrite some other
information
change the struct to TIM size of 251
|
| |
|
|
|
|
|
|
| |
usage, a matching filter should be set to allow mDNS frames to be sent to host (Android requirements).
- Fixes in BT to remove cases of
- WLAN doesn't give the antenna to the BT for eSCO retransmission (which reduce the quality of an eSCO link).
- WLAN took the antenna for two u seconds (while the antenna was allocated to the BT), it resulted in high percentage of CRC errors
|
| |
|
|
|
|
|
|
|
|
| |
Victoria Rev 6.3.10.0.137/Rev 7.3.10.0.137 - Shahar
- Fix the bug of Hard Coded TIM Element was set to 2
If we got a beacon with a longer TIM we overwrite some other
information
change the struct to TIM size of 251
Signed-off-by: Eyal Reizer <eyalr@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Victoria Rev 6.3.10.0.136/Rev 7.3.10.0.136 - Shahar
- Enable CoEx to work with A2DP SINK (taken from MCP Tree) - this is
supported for WL7 Only
- Enable CoEx to work with A2DP when the WLAN is in AP Based Role
Victoria Rev 6.5.7.0.46 / 7.5.7.46 - Shahar
- Fix bug when working in MR, and the AP role is hidden, we did not
answers to Probe Req
Signed-off-by: Eyal Reizer <eyalr@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
SR:
6.3.10.0.135 / 7.3.10.0.135
- SoftGemini - Restore of the DTIM Priority to avoid missed beacons
- fix BA session tear down when there are fragments
- Protection from including fragments into an AMPDU aggregation
- SoftGemini - Reduce BLE Priority incase of WLAN RX rate drop to 1M with Long frames to avoid WLAN Disconnect
- Remove some traces to save code RAM
6.3.10.0.134 / 7.3.10.0.134
- Fix beacons in AP mode that where program to be send before command stop role occurred
- Fix SP and tx of NULL frame in STA role while AP disappears
- Trace in basic scan to indicate whether active or passive scan is running to HEX.
- Added G trace indicates BSS Out Of Sync.
- Logger Fixes.
- Fixing the sensitivity of the device (1-2dB), on some modules, on high rates
MR:
6.5.7.0.45 / 7.5.7.0.45
- fix BA session tear down when there are fragments
- Protection from including fragments into an AMPDU aggregation
6.5.7.0.44 / 7.5.7.0.44
- Fix beacons in AP mode that where program to be send before command stop role occurred
- Fix SP and tx of NULL frame in STA role while AP disappears
- Trace in basic scan to indicate whether active or passive scan is running to HEX.
- Added G trace indicates BSS Out Of Sync.
- Logger Fixes.
- Fixing the sensitivity of the device (1-2dB), on some modules, on high rates
- Starting from this version, Trio supports only 7 peers derived from S memory limitation
6.5.7.0.43 / 7.5.7.0.43
- Fix MR GO+Sta: Recovery due to Tx stuck/Scan Complete in FW and then after disconnecting P2P Client.
- Logger Fix- Add zero mem for logger over SDIO in case memblock is full and there are several bytes remained
- Adding debug information - the option to read the process/general stack and IRQ history
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
SR:
- Bug Fix: Scan: Scheduled Scan to 16 hidden profiles is performed every 1
second after P2P Client connected to GOUT. While moving from band to
band inside a cycle, Connection scan process set its state to PENDING
w/o scheduling a timer, which might cause to a race between Beacon TX
guard interval update and Connection scan process.
Fix: Avoiding the state change to pending while moving from band to band.
- Bug Fix:protecting TimeSchedulerChangeScanToPendingCB() with DISABLE_INT()
so it won't be cut by TxNotifyTriggeredScanCB(). This is a fix for race
scenario in which we are alreay inside TimeSchedulerChangeScanToPendingCB(),
clearing the eventHandler, then cut by TxNotifyTriggeredScanCB which check
the eventHandler, and when it sees that it is 0, it moves the scan process
to pending. Then the process scheduler is invoked, moves the scan process to
running, and then we return back to the TimeSchedulerChangeScanToPendingCB()
context, which moves back the process scan to pending state.
MR:
- Bug Fix: When AP/GO is started and it is connected to STA, FW sets a guard
interval between Beacon Tx to Scan's processes of 25mSec. The motivation of
this fix is to leave a guard interval after Beacon Tx for STAs to get out
from PS to receive packets from AP. To achieve this goal scan in now is
limited to 70mSec in case AP is connected to STA.
Note: If next Scan's time event should be invoked in 500uSec or less don't
update time event.
- Bug Fix: (Process Calibration) If there is an active role tune back to the
operational channel & band after calibration completes. This cause an AP/GO
to operate on a wrong channel , i.e AP/GO transmits beacons on a wrong
channel.
- Bug Fix: Scan: Scheduled Scan to 16 hidden profiles is performed every 1
second after P2P Client connected to GOUT. While moving from band to
band inside a cycle, Connection scan process set its state to PENDING
w/o scheduling a timer, which might cause to a race between Beacon TX
guard interval update and Connection scan process.
Fix: Avoiding the state change to pending while moving from band to band.
- Bug Fix: solve an IOP issue with the Galaxy S3. We saw that when we use
AP/GO role, a galaxy S3 is connected to us, and we run TCP TX traffic,
the traffic stops after some short time. After an investigation, we saw
that the Galaxy, becomes deaf for a short time, and we try to send packets
to it but get to the station aging max retries, and remove that peer from
us which cause traffic stop. We saw that after some time the galaxy recovers
from this deafness and transmit to us again, but it's too late, we already
removed it. To overcome this:
a. In station aging, after exceeding MAX_TX_RETRIES for the specific peer
for the first time, we suspend the TX of the galaxy's flid, and
initialize a timer with ANOTHER_CHANCE_FOR_AGING_DURATION (for now 1
sec). After the timer exceeds, we resume the traffic, and try to send
packets to the galaxy again, if we get one more time to MAX_TX_RETRY
only then we remove the station.
b. There was a change in the ieee802.11n spec which allows us to transmit
BAR frames in 11b + OFDM 6/12/24 Mbps Rates we saw that it helps to
reduce the galaxy's deafness problem.
- Reducing the gLogger UART keep alive message interval from 5sec to 1sec
(UART_KEEP_ALIVE_INTERVAL)
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
- Update guard interval between Beacon Tx to Scan's processes - if next Scan's
time event should be invoked in 500uSec or less don't update time event.
- There was a change in the ieee802.11n spec which allows us to enable the
power managment bit in control frames also
- Bug Fix: When STA is connected to external AP and there is another
(external) STA connected to the same AP. The SUT will ACK all the
packets sent by the external STA.
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
| |
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Firmware X.3.10.0.131 changes:
- Bug Fix: Radio Tune was performed during TX activity and it causes the
PHY to stuck. It happens in SR, when an AP/GO's DUT is connected
to a STA/Client and run TX traffic, while STA's DUT role that is
not connected, performs schedule scan and one shot scan.
- Optimization for the guard interval between Beacon Tx to Scan's processes
solution that was delivered in Rev 6.3.10.0.127.
Firmware X.5.7.0.40 changes:
- Bug Fix: Radio Tune was performed during TX activity and it causes the
PHY to stuck. It happens in SR, when an AP/GO's DUT is connected
to a STA/Client and run TX traffic, while STA's DUT role that is
not connected, performs schedule scan and one shot scan.
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
| |
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
| |
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
|
|
|
| |
This version is based on FW logger which log record
len was removed (starting from FW X.3.10.0.129)
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
|
|
| |
- PLT FW size issue was fixed (is to be multiply of 4)
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
X.3.10.0.127:
- Bug Fix MCS00119869 - When AP/GO is started and it is connected to STA
,FW sets a guard interval between Beacon Tx to Scan's processes of 25mSec.
- Bug Fix MCS00119230 (Stop role device) - Clean HLID from HLID/FLID conversion
table on stop role device
- Bug Fix MCS00118459 (Process Calibration)- If there is an active role
tune back to the operational channel & band after calibration completes.
- Bug Fix (Trigger Scan) - When AP/GO is active , process trigger scan is limited
to 95mSec or 70mSec but scan max DWELL time remain as it was conifured by host.
- Bug fix MCS00118277 APUT doesn't send de-authentication to STA after aging time,
and perform recovery with assert number 0xc000e6e8
- Bug fix MCS00119671 Measurement Request getting stuck in FW upon receiving 0 CH
list in 2.4Ghz band.
X.3.10.0.128:
- Glogger fixes, from this version and on, glogger only works with GLOGGER version 1.7
- Added support for removing the channel switch announcement when AP preforms channel
switch and the beacon is larger then 1 mblk.
X.5.7.0.38:
- Glogger fixes, from this version and on, glogger only works with GLOGGER version 1.7
- Bug fix MCS00118277 APUT doesn't send de-authentication to STA after aging time,
and perform recovery with assert number 0xc000e6e8
- Bug fix MCS00119159 MR: Recovery 0x8a78 (_Roles_ExtractRoleContext) when operating
in MR with FW over SDIO Continuous mode
- Bug fix MCS00118268 SG SUT: Recovery (0x24d2a)(_gTrace) while inquiry or connection
in BT while TCP RX running in WLAN
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Fix MCS00117096 - If the role is sta role and we got Channel switch with count 0, then it means that the AP that
demanded the channel switch is about to switch channel at any time, so instead of 0 we put hard coded
count to 10. This value is valid by Spec.
Bug Fix MCS00118879 - SR SUT-APUT: Schedule Scan not working when APUT added as third interface
In order to fix this bug, following fixes were added:
In RemoveAllTimeEventAndRestoreSwEvents() (remove all events beside SW timers), while going through the event list, we indicate that the current time event is
SW event by adding eventType to the event info, and instead of asking about the event ID, we ask now if the type of the timer is
SW timer. Only event type other than SW type will be removed by this function
As part of this bug fix, When Starting AP role, every event in the system except of SW timers are being removed(via RemoveAllTimeEventAndRestoreSwEvents()).
Another thing done, is reseting the TSF to 0. this makes the old SW timers to be out of sync since their TSF is set to the old one. we decided
to avoid this out of sync by canceling the TSF reset and by that maintain the old timers.
Use split connection scan at all time, regardless the active role, meaning, not only during AP/GO, so we wont have to change
the scan configuration while AP role is started after connection scan is already running.
This change has also forced us to use DWELL time of 95msec for passive scan, all the time, regardless the role, and not only during AP based role.
The reason is that it was too complicated to support dynamic DWEEL time change of the passive scan in case AP role is started while passive is running.
Fix the duration set to the process connection scan. While running split connection scan, we used to set the duration
of the process to the whole scan cycle when starting a new channel. we fixed it to be set to the duration of the first channel
of the cycle because the scan is splitted per channel and so is the process.
Change the connection scan timers to be all SW timers with ID CONNECTION_SCAN_TIMER_ID (cycle timer and in_cycle_CH timer). Put all the timer handlers in the
ConnectionScanCB, and set the timer handlers to NO_EVENT_TIME when the timer call back is called.
Fixing a bug where the connection scan has registered a start pwrmgt activity at the beginning of each CH, and hasn't registered a stop pwrmgt activity
at the end of the channel scan completion, while using split connection scan. This prevented the FW from entering ELP in between the CH scan gaps.
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
| |
- When triggered scan is initiated, try to enter PS two times if the first time fails.
if the second time fails as well, do force because the chance that we lost our connection to the AP,
and the DRV didn't initiated PS. the second PS check is after 100 ms.
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
| |
Update logproxy service to the latest version.
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- P2P Provision Discovery Fix: aGO receives Provision Discovery request
from client to mac address of the P2P device.
- Adding AP channel switch for AP and AGO roles, and enable channel
switch module also for Trio.
- Reseting the Seq number of packets to something random
per when adding and reseting links
- Removing the logger dump
- Adding check to the Beacon Early Termination interrupt so if
a beacon length has been changed from prev beacon Rx,
the beacon won't be dropped in order to allow handling the new IEs.
- Dynamic SGI - Add command to Enable/Disable SGI and Greenfield
by the host for certifications tests
- Disable the assert while trying to use TRACE from NVIC level (SG DCN MGR)
- Change the recovery sequence - the driver will write 0xCAFECAFE to a certain
memory location(ADDR 0x1C), and the FW will initiate ASSERT(0) instead of
immediate recovery by the DRV. This will halt the FW priror the DRV
reads the FW core dump and 2 FW log.
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Reconfigure internal FW templates in start role for BAR to support multiple
interfaces in SR mode (for JB)
- Bug fix: trigger scan is stuck and next channel is never invoked when
trigger scan process waits to be invoked from the context of tx
exchange and the last tx frame was a management
frame->TxNotifyTriggeredScanCB
- Fix EAPOL rx filtering and handle LastSequenceNumber per link.
- AES fix: Set PN default to 0xFFFF , because some APs send the first packet
with PN 0, and if default is 0 - the FW ignores it
- Update connection monitor - protect TSFMissedThreshold from situations when
the driver set TSFMissedThreshold to 0xFFFFFFFF
- Disconnect FIX- There are pending frames stuck in TX pending queue when: BA
session is established , BA_Expected =1/0. we suggested a fix for this
issue as described below:
the cause of this issue is when the host sends stop role command then
the FW tears down all the current BA session and clears any traffic
still left in TX queues. So during the disconnect process it calls to
Links_RemoveFLIDTxQueueIfTraffic function which marks the link as
pending removal and update the callback
processDisconnect_LinksRemovalCallback to be called once the link is
finally removable , that will finish the disconnect process.
But I saw that the number of frames for the current FLID had never
been zero at the stuck point , and so the callback will not be called.
The disconnect process should not be as I have described above. the
Links_RemoveFLIDTxQueueIfTraffic function should always call to the
callback whether the flid has traffic or not, and just to mark the FLID
as invalid.
- Keep alive fixes:
1.Remove the counter from the code if a frame is stuck in the FW
queues, the Tx watchdog (or Tx monitor) will issue and assert.
2.In the tx exchange callback, not to consider a frame that exceeded
retries or timed-out as a successful transmission indicating keep-live
to the peer.
3.Any frame that passes successfully resets the time stamp sampling
for the purpose of next keep alive.
4.Continue keep alive frames transmissions while they are not enqueued,
not successfully transmitted, and no other frame is successfully
transmitted.
- Increasing BAR Threshold from 10 to 50.
- Disable PHY Clock Gating in export_hdk.c.
- Split the connection scan while in AP mode so beacons will be transmitted in
between the connection scan.
- Enable Cortex Gating during AP/GO roles. The purpose of this functionality
is to gain lower Cortex current consumption while no real activity in
FW. Meaning, when FW is just running in main with no real
interrupts/processes pending, we will allow the Cortex to "go to sleep"
until any new interrupt will be invoked (HW interrupt, timers, etc).
- Adding mDNS Filter
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
6.3.9.0.120/ 7.3.9.0.120
- During sliced scan, before every scan slice, we check if we have frames to transmit, if so we should transmit frames for 50 millisecond and then
(or until we are out of frames to transmit) move to the next scan slice and continue the scanning.
Before the fix, for every single time we check our queues of frames to transmit, we get a negative answer and continue scanning without giving
a chance to the frames to be transmitted and released.
- Fix for Bug MCS00116357- WiFi TGn APUT 4.2.31: The APUT doesn't turn on the HT non-member protection bit
Original Bug issue was raised in Bug MCS00113136. MCS00113136's Fix did not handle a stop start scenario.
- Reconfigure internal FW templates in start role for BA to support multiple interfaces in SR mode (for JB)
6.5.7.0.32 / 7.5.7.0.32
- Update MAX_STA_BASED_ROLES to support 3 sta based roles (instead of 2).
Driver uses additional interface for the P2P connection establishment which requiers the third sta based role enabled.
- During sliced scan, before every scan slice, we check if we have frames to transmit, if so we should transmit frames for 50 millisecond and then
(or until we are out of frames to transmit) move to the next scan slice and continue the scanning.
Before the fix, for every single time we check our queues of frames to transmit, we get a negative answer and continue scanning without giving
a chance to the frames to be transmitted and released."
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
SR:
- Reconfigure internal FW templates in start role to support multiple
interfaces in SR mode (for JB)
- SDIO Logger support, Old protocol alignement and GTRACE enabled by default
- Bug Fix: Current Consumption in PLT version
MR:
- SDIO Logger support, Old protocol alignement and GTRACE enabled by default
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
|
|
|
| |
wl_logproxy was updated to latest ane greatest version.
thanks to Gil Barak.
Signed-off-by: Yair Shapira <yair.shapira@ti.com>
|
| |
|
|
|
|
| |
FW logger over sdio is now being forwarded via socket over adb
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Victoria Rev 6.3.9.0.118 / 7.3.9.0.118 - GuyM
SR:
- Increase the Role DB to 5 records (Each IF enables DEV and STA roles).
- Support scan (DEV role enabled and not started) while AP/GO is running using
SR FW by not attempting to enter power save.
- Added mechanism to allow AP/GO beacons to be transmitted in between passive
scan.
- Increase the AP/GO Beacon priority to 3.
- rx_complete_1273.c: remove "nokia camp" patch in order to not drop frames on
the second interface.
- IEIterator_GetNextIe function bug fix: Check if the remainingSize is valid
(the remainingSize cannot be 1 or negative). We do it to protect of frames
that don't have CRC error but have wrong data at the IE section.
- host_slav_manager.c: Allocate an extra memblock for FW generated frames also
for Data frames.
- Update the usage of the connection scan - in function
ConnectionScanScheduleNextCycle use the value pointed by the
intervalIndexToUse instead of intervalIndexToUse itself.
- Bug Fix MCS00112566 - WiFi Direct: Recovery occurs after successful
group formation. PowerSaveModule_ExitPowerSaveByClient() - function
callback points to null!
MR:
- Update the DISABLE_PV macro - make sure that the DISABLE_PV operation is
atomic. Therefore we added a READ_PV() action in order to have HW ack that
program operation was canceled successfully.
- Init the frame control when init the QOS NULL Template (setTXQOSNulTemplate)
- Free the security context of the broadcast keys for AP based roles.
- Bug Fix MCS00115363: Schedule Scan in Multi Role is acting different than
in Single Role.
- Added Suspend role for AP/GO:
1. Added mechanism to suspend/resume Beacons.
2. Update the MRS logic so that STA Based roles will have priority when
choosing the operation channel.
- IEIterator_GetNextIe function bug fix: Check if the remainingSize is valid
(the remainingSize cannot be 1 or negative). We do it to protect of frames
that don't have CRC error but have wrong data at the IE section.
- host_slav_manager.c: Allocate an extra memblock for FW generated frames
also for Data frames.
- Update the usage of the connection scan - in function
ConnectionScanScheduleNextCycle use the value instead of the index.
- Bug Fix MCS00112566 - WiFi Direct: Recovery occurs after successful
group formation.
PowerSaveModule_ExitPowerSaveByClient() - function callback points to null!
Signed-off-by: Ido Reis <idor@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
-Fix for Bug MCS00115194: APUT not responding to Probe Request in rate 5.5 Mbps
APUT dropped the probe request frame in the case the frame didn't include PBC on.
This is an update for Bug MCS00111501 - No need to drop the probe if it doesn't include PBC on.
-Fix for Bug MCS00115195 - SUT not responding with EAPOL response until NULL Data is transmitted while connection with WPA2 PSK security.
Problem was that (mainly in case of TI-TI) the first EAPOL frame was dropped by the SUT due to EAPOL action filter, when there is match of
lastEAPOL seq to the one stored.
In this fix we initialize the seq num of last received EAPOL frame once EAPOL exchange is completed.
Seq num is 12bits long, meaning, the init value we used is of 0xFFFF and can never be matched.
We init the value at the end of the EAPOL handshake (after connection is established) for the next EAPOL handshake (next connection)
since current handshake is over and there is no need to the current seq number value until the next handshake.
-Update the DISABLE_PV macro - make sure that the DISABLE_PV operation is atomic. Therfore we added a READ_PV() action in order to have HW ack
that program operation was cancelled successfully."
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Service Pack 1 Bug Fixes:
- Fix for Bug MCS00112750: Scheduled scan stops after 12 mins & 45 secs.
- Fix for Bug MCS00108221: During BT discoverable, when exit ELP, the BT InquiryScan activity set to 1, causes Wifi scan failure.
- Fix for Bug MCS00112788: Add action to rx filter - add the possibility to set the action dynamically and differently for every filter.
- Fix for Bug MCS00112837: Increase num of Rx filters to 7 (#define MAX_DATA_FILTERS 7)
* From 2.114:
- During opening a tx BA session, if the recipient returns ADDBA response with winSize smaller than
we requested, we don't update it properly upon the TBC_Init.
- rx_streaming.c: Go over all the TIDs (8) instead of RX_STRAMING_NUM_OF_TID_SUPPORTED (4). Also send the TID instead of the TIDindex.
- RogBug Fix MCS00113136 - WiFi TGn APUT 4.2.31: The APUT doesn't turn on the HT non-member protection bit
- Eli: Bug Fix MCS00110974 - APUT : Beacon stops while UDP RX traffic - Wrong AIFSN calculation.
* From 2.112:
- Added two more Probe Request types for application scan (one shot), and changed the
Older Probe Request templates to Connection scan (sched scan) probes. also changed
the probe to be regular and role aware and role protected.
* From 1.111:
- Bug Fix Mcs00111501 Wi-Fi Wpsv2 Aput 4.2.13:The Aput Connected When Overlapped Pbc Sessions.
Fixed several issues in generic_ie_parser and anded support for pbc in probe response
1. In function FindAttribute fixed an issue where the while was stopped after one iteration.
2. In function FindAttribute fixed an issue where the pointer was increased only by attribute
payload length without the header length.
3. Added support for IE that uses BIG Endian.
4. Added function that search and indicates whether the PBC was pressed the probe response WSC IE
and we use that function to decide if to raise the probe to the host or drop it.
* From 1.109:
- Bug Fix MCS00111486: PS/ELP: ELP doesn't work anymore after re authentication with Cisco in mix UPSD mode.
- Bug Fix MCS00111349: Power save: The SUT doesn't send trigger pkt (QoS Null Data) after roaming.
- Bug Fix MCS00111892: Wi-Fi 11g APUT Certification Test 4.2.3.6 Fail.
- Bug Fix MCS00112301: SDIO crash when running BT A2DP during Wi-Fi connect/disconnect."
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
| |
- Bug Fix MCS00111656: Current Consumption: high cc when the SUT is in Power Save mode, and connected to AP.
- Revert the power consumption improvement for AP mode in suspend where we do not forwarding to host probe requests containing WPS IE (done in 6/7.3.5.0.99)
as it is failing Wi-Fi WPSv2 APUT 4.2.13 test (The APUT connected when overlapped PBC sessions)
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Release notes for version 110:
- Partial fix for bug MCS00107046 - Stability APUT: TCP traffic
stopped after 10h. STA still connected. Tx stuck. New STA
fail to connect.
After delba caused by expiry of BA session we didn’t do
tear down to the session and in some scenarios the
fw queues where still loaded with traffic. A teardown
to clear the queues was added.
Signed off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
|
|
| |
Release notes for version 109:
- Bug Fix MCS00111551: If Driver set the EnableRate to Zero, FW use 2.4 low rate (1 Mb/S).
- Bug Fix MCS00111621: Add additional filter decision (dsps_decision) to rx_filter.c in order to filter out probes/beacons from different channels.
- Bug Fix for MCS00111628 ELP: HP FEM Trio remained awake (never enters ELP) after BT was turned off.
- Add Assert under GTRACE_ENABLE flag for invalid role id in Roles_ExtractRoleContext.
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
X.5.4.0.21
Release notes for version 108 (single role):
Added RX Reset for each TX Frame that is not immediate response.
Adding Chicken Bit (BIT[9]) for ECO in the HW that will drop frames with FCS Error before sent to the FW (RX complete).
TxMon 0x400 fix:
1. BA was not handled when immediate response for this BA was requested.
2. BA arrived after a long time and caused the clearance of the pending list while BA expected stayed 1.
API was changed in previous version (Supporting 3 RX BA Sessions). The Version number was updated in this version.
Release notes for version 107 (single role):
- Fix for OMAPS00267085: When connecting to a Hidden SSID AP we keep the host awake. The problem was that we did not
filter correctly beacons in BeaconFilter_Action() and we sent ALL the beacons to the host. This might happen
in case we have received beacon with multiple SSID IEs for example.
- Supporting 3 RX BA Sessions as SUT n order to deal with IOP issues of APs which might open three BA
sessions (e.g DIR615 which requests BA on TIDs 0,1,2).
- Adding RX_RESET at TX_Start as a W/A for the TRIO ACK-on-ACK HW bug.
- In TxEngine_CancelProgrammedFramesForFLIDMask() - After disable PV we can still get Tx start after we read the int
status. The fix is kind of a small delay by reading the value from procedure instead of pointer.
- HassanH: Bug Fix MCS00109757: Dynamic PS: STA does not enter to PS mode after burst of ARP Request with the STA's IP address
- Adding RX_MON mechanism - detecting cases of more than 2 seconds of no free rx memblocks, and initiating recovery.
This was seen at Amazon and caused to download stuck.
- Adding enable of two chicken bits in HWfor Stomp&GO.
- Configuring the min delimiter size to 13 bytes instead of 1 byte.
- Reverting SW S&G feature and enabling back the Aruba Fix (HW S&G).
- Adding partial W/A for the Trio ACK-ON-ACK HW bug under the define MISSING_IMMEDIATE_RESPONSE_WA.
The Trio ACK-on-ACK HW bug W/A is enabled only in Trio!
Release notes for version 21 (multi role)
- Supporting 3 RX BA Sessions as SUT n order to deal with IOP issues of APs which might open three BA
sessions (e.g DIR615 which requests BA on TIDs 0,1,2).
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
FW comments:
- ShaharP. - Full Stomp&GO feature support
- Fix bug MCS00104692 - Bss loss timer (BssLossEvent) and
TSFOutOfSyncSwTimerSet don't clean in all cases
- Fix bug MCS00105324 - QOS - AIFSN Registers are configure incorrctly
- Fix bug MCS00110721 - IOP: BSS loss when working with Belkin AP due to AP TBTT glitches
- Fix bug MCS00110423 - IOP issue with APs not receiving seq num 0 during EAPOL handshake
- SG:
- In checkCoexAntennaState() - avoid false debug traces of
SG_IMP_ELP_ANTENNA_MISMATCH when WLAN connected in 5GHz
- Adding FRAME_TYPE_REASSOCIATION_RESPONSE condition in Rx_handleRxCompleteFrame)
and in TxPreXfr() to secure SUT connection and remote STA connection
to APUT during ACL.
- uint32 alignment of two parameters in function BeaconFilterCalculate32Bit()
- Increasing the TX MON timeout to 2seconds
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
X.5.4.0.20
FW comments:
- Update the p_WasHwTsfUpdated flag only after checking if Beacons are from our AP.
- Handling false Ap Jitter scenarios (phase I of this complete solution).
- Code Reduction and Code Optimization related to IsPGVersionTrioPG2 in rx_complete_1273.c.
- Bug Fix MCS00109245: STA doesn't enter ELP - High power consumption. We stopped the wrong activity by mistake.
- Bug Fix MCS00109088: FmCoexInit function with wrong parameters.
Wrong configuration of TOP_FREF_26_MHZ/TOP_FREF_52_MHZ in fm_coex.c.
- Bug Fix MCS00109373: Fix for the wrong FLID during BA session.
- Bug Fix MCS00109084: IOP: SUT BSS loss in open air TCP Traffic to Aruba AP (WorkAround since root cause was not found yet)
- Added FW dump debug tool is main.c.
- Bug Fix MCS00109310: SUT: Add protection in FW in case of large DTIM periods.
- (MCS00108403)(MCS00108403) - Fix command complete timeout errors (mostly stop role). This happened because we did not
call ENABLE_INTS() in SoftGemini_StopWlanActivity() and RxEarlyTermNotifyBeaconDtim()
- Added a newer version of dis470 to cgtools. All compilations will now include a disassembly of the entire fw.
- Bug Fix MCS00107873: MR: During STA traffic and p2p find in second role, Recovery may occur.
Update the MRS to execute the ROC command serially. The host won't wait to complete event on second ROC.
- Remove CFG_RX_ALL_GOOD filter from RX_CFG_OPTIONS_ROLE_COMMON.
- Increment the pointer in chain.offset and not in (chain.mblkPtr + chain.offset) in function CopyFromMemBlockChain.
- When we got a prob-res that wasn't match the SSID that we looked we set this frame as not success.
The function SetRxFiltering was suppose to recognize this case and drop the frame but instead we pass the frame to the
host because it was linked to the system flid.
- Code Reduction and Code Optimization related to IsPGVersionTrioPG2 in rx_complete_1273.c.
- (MCS00108403)(MCS00108403) - Fix command complete timeout errors (mostly stop role). This happened because we did not
call ENABLE_INTS() in SoftGemini_StopWlanActivity() and RxEarlyTermNotifyBeaconDtim()
- Added a newer version of dis470 to cgtools. All compilations will now include a disassembly of the entire fw.
- Bug Fix MCS00109084: IOP: SUT BSS loss in open air TCP Traffic to Aruba AP. (WorkAround since root cause was not found).
- Added FW dump debug tool is main.c.
- Bug Fix MCS00108513 SUT: doesn't enter to ELP after almost every connection to AP.
- Bug Fix MCS00109310: SUT: Add protection in FW in case of large DTIM periods.
- Bug Fix MCS00109373: Fix for the wrong FLID during BA session.
- Bug Fix MCS00109088: FmCoexInit function with wrong parameters.
Wrong configuration of TOP_FREF_26_MHZ/TOP_FREF_52_MHZ in fm_coex.c.
- Fixes for Bug MCS00108095: MR FW: Second role SUT won't connect to ext-AP if scan was not performed beforehand.
off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
| |
X.3.7.0.103 TP degragdation fixed
X.5.3.0.19 revert PPDU_WA
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
| |
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
| |
Signed-off-by: Igal Chernobelsky <igalc@ti.com>
|
| |
|
|
| |
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
| |
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
| |
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
| |
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
| |
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
| |
corresponding fimware logger files.
Signed-off-by: Assaf Azulay <assaf@ti.com>
|
| |
|
|
|
|
|
| |
Change the firmware names to follow the new naming convention, update
ili file names and Android.mk files.
Signed-off-by: Luciano Coelho <coelho@ti.com>
|
| |
|
|
|
|
|
|
| |
Align the MAC address argument with the one use in set_nvs. With this
patch we can get the MAC from fuse, use the default or pass a specific
address to the autocalibrate procedure as well.
Signed-off-by: Luciano Coelho <coelho@ti.com>
|
| |
|
|
|
|
|
|
|
| |
We will need to keep the power on after doing BiP in order to read the
MAC from the fuse (and possibly other features in the future). Thus,
we need to move the power_on/power_off calls out of the
plt_do_calibrate() function.
Signed-off-by: Luciano Coelho <coelho@ti.com>
|
| |
|
|
|
|
|
| |
Generating random MAC addresses is evil, and even breaks IEEE rules if
the OUI is also randomized. Don't generate random addresses anymore.
Signed-off-by: Luciano Coelho <coelho@ti.com>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The default MAC address is 00:00:00:00:00:00, which means that the
driver will read the address from fuse.
This is redundant with manually assigning the address to zeros, but it
checks whether the hardware supports reading the MAC from fuse ROM and
fails if not. This brings more awareness to the fact that the MAC may
be unavailable in the fuse.
Signed-off-by: Luciano Coelho <coelho@ti.com>
|
| |
|
|
|
|
|
|
|
| |
If the NIC part of the MAC address assigned to the NVS is FF:FF:FF, we
will wrap around in the driver and use 00:00:00 for the second
interface. Add a warning to the calibrator tool so the user is aware
of this.
Signed-off-by: Luciano Coelho <coelho@ti.com>
|
