Age | Commit message (Collapse) | Author |
|
Colin (ccross@google.com) found out that create() gets
called on same object multiple times; without a call to
destroy() in between; this causes leaks.
This change will destroy the request object, before allowing
create to proceed.
BUG: 25392275
Change-Id: Id2199346020dd69f735f50fe3b2f11bfe251cad0
|
|
The commit "Add the fw diag msg hdr" introduced an error
while calculating the offset to the next buffer. This
caused incorrect firmware log entries in the ring buffer.
Change-Id: Iccce94c33a41f62f4e7c3477f603a6c8d33a1efa
CRs-Fixed: 913224
Bug: 24350330
|
|
In earlier log cleanup few logs got missed, cleaning up the same.
This change involves cleanup of wifilogger module as well.
Bug: 24266530
Change-Id: I270a9d295e04938df75a5f15f38d3cfc05360a5b
|
|
Member variables related to bssid hotlist ap lost were not
initialized resulting in invalid values. This change corrects
it by initializing this appropriately.
Bug: 24266530
Change-Id: I686cb43c860e8f61682d8276461cd7f204dc19b8
|
|
Fix incorrect pRsp assignment for the getFurtherAvailabilityMap.
Add a check to verify the length is always greater than the
TLVs passed as response.
Bug: 24066021
Change-Id: I987deded13c3dae93fe81dd8a9e4b96b73fc1eff
|
|
There was some changes done in NAN files to fix compilation
error which were incorrect. Correcting these appropriately
as part of this CL.
Bug: 24066021
Change-Id: I1131a8be34f9dc6318c98005a05f958214800fb2
|
|
Cleanup unnecessary logs from llstats, gscan, rssi monitor
common and tdls modules.
Change-Id: If66a8f61b11a101d84546a06514aeb5e0d8fcbfa
Bug: 23942132
|
|
Buffers will be read if ring buffer reaches minimum configured
threshold. It is possible to reach minimum threshold after writing
header to the buffer and before writing actual payload. Incomplete
buffer will be read out in this case. This commit ensures to read
the data from buffer only if complete record is available.
Change-Id: I0c558e4cf73ea31892c8c086f2310405ebc49c74
CRs-Fixed: 904848
Bug: 23934884
|
|
In case of driver prints and firmware prints, events will be
split into write to ring buffer as header and payload. Write
header and payload to same buffer so that complete record
will be available in a single buffer.
Bug:23783279
Change-Id: I54d3f813fbb14932b483cd6423d41c7187a464a9
|
|
Currently, ring buffer writes and reads happens with byte
boundary. Add support to write only records and do not split any
records.
BUG=23783279
Change-Id: I2342f86cc4c3a678f21f1595b81e0cdf6d24b2c0
|
|
Ring buffer write fails if the ring buffer is full. Push the ring
buffer data to upper layer and then try writing the same data
again to the ring buffer.
Bug: 23758463
Change-Id: I3c9c310bec43d60c580af43157ca174aa524f378
|
|
Include the fw diag msg hdr as part of payload being
passed to ring buffer. The firmware logs require
the diag hdr info while parsing the logs collected
as part of the bugreport.
Bug: 23758463
Change-Id: Iacdfc0d380990bb6b1dc7acf8edf95a2302abb63
|
|
As part of earlier commit log_handler and alert_handler were
protected by mutex. However, calling framework callback within
mutex context would cause deadlock, thus unlock mutex before
calling the framework callback and reacquire again when required.
Bug: 23530883
Change-Id: If4cd92be906307a9e5ff7a8002b219f93f1bda64
|
|
If an already registered vendor event tries to register
for the second time, update the call back function but do not
add a new entry to the handlers.
Bug: 23483365
Change-Id: I93446721b2e75b768c8f7b40583e7ac7d1e5c282
|
|
Treat non zero as error instead of checking explicitly for
ret < 0 only.
Bug: 23483365
Change-Id: Ic3a331b323eede958cf811086e1f2d7e0cfb2ff7
|
|
RSSI monitoring gets issued from multiple places and request id
can be different. So, no need to consider request id for
deregistration.
Bug: 23483365
Change-Id: Idf93098e0273b6afaff161d85d805a317714dc24
|
|
We are seeing a race-condition when the event handler function
is getting events from the driver and at the same time cleanup
is also happening from the command side. Due to the race-condition
we are seeing a crash. This commit addresses this by ensuring that
the cleanup of event handler happens only when wifihal cleanup is
happening and also would happen in the same context in which events
are processed.
BUG: 23277168
Change-Id: Ib0cbcbd23e2c0424d6f81597685d085d95df340d
|
|
Querying for valid channels immediately after country code change
may result in invalid channel list as kernel takes some time to
populate the channels in driver. A little delay(50 ms) between
set_country_code and get_valid_channels can solve the issue.
Bug: 22937052
Change-Id: If252ca9512be1cbc9d0b3e4fc80cb703657fc6f2
|
|
Protect log_handler and alert_handler with mutex as these
can be accessed in both command and event paths.
Bug: 23292340
Change-Id: I0565d818e97f81609152b288cc9074d995232744
|
|
Per packet stats module collects below stats from first MPDU of
the AMPDU.
1. RxMCS
2. last_transmit_rate
3. rssi
4. tid
But the first MPDU may not be received by
hardware/corrupted sometimes. Populate the cached stats
from first MPDU of the previous AMPDU in such cases.
Don't parse Rx packets if the REORDER flag is set as parsing of
these packets may show duplicate entries in the stats.
Change-Id: I8cf78f42e055b0c22d123f21c91f2c204087afe0
|
|
Firmware sends valid TID only for first MPDU. Populate the same
to all other MPDUs.
Change-Id: Ie06b1e032c105c3bcd2ccf3ddf9c92524b71c518
|
|
Frequent updations are expected to the stats structures from
firmware. Validating the events received based on its exact size
may not be correct in this case. So, consider events as valid only
if the size of events is not less than expected.
Rx peer info event has been handled with the help flags already.
So, remove the duplicate check from parse_rx_stats()
Change-Id: I42f2a3f402684fd3de049d7d9fe011068fabb45b
|
|
There were few handlers missing initialization in the function
table which are added as part of this commit.
Change-Id: I136879fd6471b5d424ba86fbe085a66313bc2d6a
|
|
Usage of rand and srand function is a moderate security
vulnerability, thus removing the usage of this and replacing
with arc4random.
Change-Id: I5f24e96817bf3de09848f5626278ac65b00e07ed
|
|
Earlier wifi_get_firmware_memory_dump was processed asynchronously
resulting in the event getting dropped. Resolving that by processing
that synchronously.
Change-Id: I93a4bb9a22d9de98c43ecc93b792fe47785b1cfa
|
|
Lowi module initialization is causing crash and this happens
as part of gscan functionality. Since, Lowi is not processing
GSCAN functionality now do not initialize Lowi module to avoid
the crash.
Bug: 23004442
Change-Id: Ibadd2f78eb2511054c671fa7434ee3d024e45fc9
|
|
In case of report_event being 2 each scan result is forwarded
resulting in lot of logs, reducing the logs for the same.
Change-Id: I0002939e529fb1eb9bc03895efd3bc4d3fcb64c6
|
|
This commit removes the global usage of various Gscan event
handlers and makes it part of hal_info structure so that it gets
cleaned up as part of wifihal cleanup.
This commit also addresses the appropriate cleanup as part of
wifihal cleanup.
Bug: 22504320
Change-Id: I56ea636e6d47b5fcdd66a9e6db4e2a36e8dac8ba
|
|
Add support for GSCAN CAPABILITIES event
as part of the Debug Framework.
Change-Id: I8c90e56637171a9719f97dc112042037a600e79e
|
|
Add reset_alert and reset_log handlers to function table to
facilitate the upper layers to call these APIs using the
callbacks.
Change-Id: I33e40a5ffe563e1dd19f63287270a8031bb35a64
|
|
Get valid channels is independent of gscan feature. Hence remove
the dependency.
Change-Id: I33d6196722a27aea19053afefadb9f670baf1a25
|
|
Add an attibute to send channel list in nested format. This
attribute can carry information about multiple channels.
Scan interval is to indicate interval between PNO scan cycles
in msec.
Change-Id: I86d9520dc09573676bca000c627d73b6fef8dbc3
|
|
try_status has a parameter "series" which needs to be
considered to select the appropriate series_bw structure.
Change-Id: If4cb8c4e54dccf60d083dcaf8ea44a4e51c09046
|
|
It was observed that during reportevent 2 case the scan status
event was lost since there are multiple back to back fullscan
event coming. This commit increases the buffer size and also
removes some unnecessary gscan logs.
Change-Id: Id35ef90dc435ec9d4a0d56c8bfc20d02eb7b8abc
|
|
Remove the redundant logs and print stats in a structured
format with ALOGV log level.
Change-Id: I65e604717389e72998924156089c0e5f1d4f2e42
|
|
1. Firmware sends stats for failed Rx packets with corresponding
error code. Set Rx packet status as success only if none of
the error bits are set.
2. Indicate tsf timestamp as the packet timestamp for Rx packets.
3. Correct the rate index calculation.
Change-Id: Icbe0c0bb8d8d61d55d1f01b247eb784365dcf8f4
|
|
Process only relevant Rx events coming from driver/firmware to
generate Rx per packet stats.
Change-Id: I3e3aa309f0cfee828f8075a78f8ff47f1f6da37a
|
|
Get the bandwidth value from Tx stats try list based on total
tries and calculate rate based on its value.
Set the RSSI value for failed Tx packets to 255, which indicates
an invalid value.
Change-Id: I7c4b43268627eef1ae6576bd6afd37f1bf55c630
|
|
Decode short gi value from Rx stats and correct rate calculation
based on short gi value.
Change-Id: I5db98d3363298820095ee3d0bcd809745b935a4f
|
|
Firmware sends stats only for the first msdu of the AMPDU.
Populate stats for other packets based on the tx packet bitmap
and block ack bitmap.
Change-Id: I530fccc251228962b105d9c18961d29777417d82
|
|
There is no need of returning error code if the ring buffer write
fails because the ring buffer is not enabled yet. Hence donot
return error codes in these cases.
Change-Id: I8518f341c0057b24cf97e8d63592cb8b034422c4
|
|
Fix wrong pointers passed to addLoggerTLV() to avoid possible
page faults.
Change-Id: I70b67f950d9036199ff13901d44285cfc250e1a2
|
|
While setting on_alert handler, check should be for non-null
handler but not null handler.
Change-Id: I1823e72cf6d05779f1cfd018fdb5c69a5f570187
|
|
Aggregated Rx packets may have valid data in a specific MSDU of
the MPDU. We need to cache the stats received so far and update
once the expected MSDU received.
Below stats for Rx aggregated packets contains valid values only
in the first mpdu of the AMPDU.
1. RxMCS
2. last_transmit_rate
3. RSSI
and timestamp is valid only in the last msdu of the last mpdu.
Write the updated stats to ring buffer.
Change-Id: Ie09992926ae447995946b64c3e8e07dd0650cd5d
|
|
We may need request id in WifiVendorCommand class to compare with
the request id received in asynchronous events. Maintain the
request id received from invoker in the object of
WifiVendorCommand
Change-Id: I6514c912dc41212dc22703af11220378d19551bc
|
|
Roam Auth complete event is incorrectly mapped in Wifi-HAL,
remapping the event to match the driver.
Change-Id: If627ce33afc73513be0ae73e7105f4777968a1d5
|
|
The ASSOC complete event is incorrectly mapped. This event is
generated when the connection is complete.
Change-Id: Ice68f13b9748c617238a22feea3ed2ec5d08460f
|
|
Include the Channel Switch anouncement event as apart of the
connectivity events.
Change-Id: Id702940ece7aa2d2f94e1c05c314887b9ba2a93f
|
|
Add a new attribute to notify the driver whether to use the
current EPNO request as PNO or EPNO. Driver can take decision
based on its value.
Change-Id: I1c8b05fb6c555ecf436ec5832ce0a4ea92c8aec0
CRs-fixed: 855802
|
|
LCI/LCR related changes
Change-Id: Ib5f2cf83bbfeb041e4f6fd4302dd9a45af2e0284
|