wdi20 enable implementation for hsoffload. Large physical address
support and wdi20 interface propagate to IPA.
This is part of HIF and DP.
Change-Id: If898f03a8a75689c98c70d1e55910f476d8fe329
CRs-fixed: 941041
Target_lock was a global lock for all target and copy engine data
structure access. Replacing target_lock with a per copy engine
lock permits paralel completion handling for different copy engines.
Change-Id: I9c09d557c85f9e48beafe5e3f936e105183ddd3f
CRs-Fixed: 941355
This is called by the platform driver in case of kernel panic.
Replacement syncronization might be needed here if the kernel
panic doesn't stop the other host cpus before the platform driver
calls this.
Change-Id: Ic6d4eb1cb08f6363029463994ec7e3837212e271
CRs-Fixed: 941355
If watermarks are going to be used, the should be set as part of a
single threaded initialization sequence.
Change-Id: I3e1e9cd0f55f46d9f46718c139963931311fe8ae
CRs-Fixed: 941355
No other thread should modify the copy engine indexes while we
are cleaning up the data structures.
Change-Id: I324286c3a8eed24ddd94aa868dbaac425cc30dbf
CRs-Fixed: 941355
Remove ce_disable_any_copy_compl_intr and
ce_enable_any_copy_compl_intr since they aren't
used and give a false sense of thread safety.
Modifying copy complete interrupt currently needs
to be synchronized with completion handling.
Change-Id: Ieb27a8c71363ebd56f16eda9e1037a161a5eeb62
CRs-Fixed: 941355
Send and recv callback registration currently occurs durring init
and after bmi download is complete. In either case no transactions
are in progress and no race condition can exist.
Similarly watermark callback registration would happen at the same time
if it were used.
Change-Id: Ia77bf30fb3bcefa500465cea9ad9927990bb7ced
CRs-Fixed: 941355
Download length should be set before messages are recieved, so
no race condition should exist.
Change-Id: I92de4b3c4077f23032df45ceca6e169616351b8b
CRs-Fixed: 941355
FW is asserted, dont service the messages. The hif completion
thread used to prevent processing of messages when firmware crashed.
Change-Id: I42a5a285056e989d302e766f2922bb7c114634dc
CRs-Fixed: 941358
ce_enable_msi should happen before interupts are enabled. Therefore
we should be single threaded here and don't need to protect access
with a spin lock.
Change-Id: I24feec02003bdb402201d71d1cd5254dacae0b72
CRs-Fixed: 941355
ce_init should happen before interupts are enabled. Therefore
we should be single threaded here and don't need to protect access
with a spin lock.
Change-Id: I5e1cfd32860eeda8a403744cf1589e35c57ee6df
CRs-Fixed: 941355
We want to process the completions in parallel.
Upper layer should take care of its own thread safety.
hif_completion_thread is a legacy function that does in
context serialization of hif completions. Bypass it.
Change-Id: I7953545b55dae093931741359bbf985e2fb9faca
CRs-Fixed: 941358
Propagation from qcacld-2.0 to qcacld-3.0
- Current AGC registers dump logic is not correct and its not
reading correct target registers
- Add correct AGC register read recipe to read correct AGC data
- Wake up SoC before reading any target register
Change-Id: I38016f91bf3d124e93e304522058e97daa711f4a
CRs-Fixed: 911554
Some of these are very remote scenarios (like HIF not being
initialized at the time of the calls, which could then
dereference a NULL pointer). There are a few cases of looping
beyond boundaries.
The scope of this update is limited to NAPI related code.
Change-Id: I60229406d5ab514d5465ef2569324d3d4eb430d4
CRs-Fixed: 938253
We want to process the completions in parallel.
Upper layer should take care of its own thread safety.
hif_completion_thread is a legacy function that does in
context serialization of hif completions. Bypass it.
Change-Id: I73b8518d87645d827814cd1345eb68370de30d1f
CRs-Fixed: 941358
We want to process the completions in parallel.
Upper layer should take care of its own thread safety.
hif_completion_thread is a legacy function that does in
context serialization of hif completions. Bypass it.
Change-Id: I19660327ebb4a3cbb9908ce562ab6c72a6282e95
CRs-Fixed: 941358
qcacld-2.0 to qcacld-3.0 propagation
In athdiag_procfs_remove, set proc_dir to NULL after removing the
directory. This is to avoid trying to remove the directory which is
already removed.
Change-Id: I01526f4b59ccc991d157c1028d4df5904338f182
CRs-Fixed: 868965
The SRRI/DRRI can be read from DDR instead of doing an
actual hardware read. This mechanism is compiled out by
default.
Change-Id: I8fc6c8817b5958f3d00b47ccd3dba36b354097e8
CRs-Fixed: 932344
Facilitates faster data transfers. This provides a basis
to start using the other 4 CEs for additional services.
Only adding structure elements to enable these CEs not
really using them yet.
Change-Id: I4be07413de409195ccc717765c75806a1470941e
CRs-Fixed: 938752
- Fix a crash seen due to memory being freed twice
- Deregister the LRO flush callback during LRO disable
CRs-Fixed: 933123
Change-Id: I18edadc620da04b9a571da6e1b143c099c308d90
cds_get_bin.h and cds_get_bin.c were derived from qcacld-2.0
vos_get_bin.h and vos_get_bin.c. These files got their names
since at one time they housed the wrappers used to retrieve
binary files nv.bin, cfg.dat, and firmware. But over time the
functionality to actually retrieve binary files (get_bin) has
been removed, and almost all that is left is concurrency
management functions. So, moving the .[ch] functionality to
the right files.
CRs-Fixed: 932777
Change-Id: I4a26c304cc6b6224d0839c365e4006fd32270961
qcacld-3.0: Initial snapshot of ihelium wlan driver
to match code-scanned SU Release 5.0.0.139. This is
open-source version of wlan for next Android release.
Change-Id: Icf598ca97da74f84bea607e4e902d1889806f507
