| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|
|
|
|
|
|
| |
corrected the structure value to be update : so that the
Comp returned Value will be notified to the APP
Signed-off-by: Radha Dasari <radhapurnima@ti.com>
|
|
|
|
|
|
|
|
| |
code clean up for domx, removing dead code
Reverting back omx_rpc_stub.c, timm_osal_pipes.c and
omx_proxy_common.c for future use and reference
Signed-off-by: a kasi viswanathan <kasi.alageswaran@ti.com>
|
|
|
|
|
|
|
| |
Removal of compiler warnings.
typecastings of datatypes, adding default statements for switches, etc.
Signed-off-by: a kasi viswanathan <kasi.alageswaran@ti.com>
|
|
|
|
|
|
|
| |
Memory unmapping to be avoided for null buffers
Change-Id: I9c2a7fd1029f314f4feddaded981799de0a14a84
Signed-off-by: Saurabh Bipin Chandra <a0131926@ti.com>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Adding PROXY_ASSERT to ensure going to EXIT tag in every function
in case of error from RPC layer.
Discovered in call from PROXY_UseBuffer where even with error from
RPC we were going ahead and update buffer number and updating with
garbage pointer.
Other cases that get affected are: OMX_ProxyCommonInit, PROXY_UseBuffer,
PROXY_AllocateBuffer and RPC_UTIL_GetStride
Signed-off-by: Gabriel Guerrero Hernandez <x0099164@ti.com>
|
|/
|
|
|
|
|
| |
Certain configs (i.e. GetComponentVersion) is exceding
the current 240 byte payload and causing segfaults.
Signed-off-by: Tyler Luu <tluu@ti.com>
|
|
|
|
|
| |
This reverts commit 638b146a36bccb17ed5a4674aeb9553d5203ccf6, reversing
changes made to 72fb683338bc803f15fa9bdf950562255be6c1bf.
|
|
|
|
|
|
|
|
|
|
|
| |
Currently the Max buffers was capped to 32. But certain clips
need more than 32 buffer header.
Hence increasing to a safe value of 100.
Size of each entity is < 50 bytes. So increasing to a large value
will not have any side effects.
Signed-off-by: Sarthak Aggarwal <sarthak@ti.com>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Increasing the payload size to max allowed by rpmsg
driver which is (496 bytes - 12 bytes for omx header)
Certain configs (i.e. GetComponentVersion) is exceding
the current 240 byte payload and causing segfaults.
Change-Id: I535b016b6e30dea53c19c8745a471f19e72ee011
Signed-off-by: Tyler Luu <tluu@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Change-Id: If2cb2f7744967df6632dd097bd5a9576cf4f9718
Signed-off-by: Sarthak Aggarwal <sarthak@ti.com>
Conflicts:
domx/omx_proxy_common.h
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This patch enables to map an array of 2 buffers instead of a single buffer
in SetParameter() by indicating the number of "local" buffers that need
to be mapped to remote core. These 2 buffers must be an array. Exiting
camera proxy is updated to reflect this extension.
Change-Id: Id939dc8a69e693d1ff7c627771960f1bb3ffc946
Signed-off-by: Abhishek Ranka <asranka@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Description# setprop controlled mechansim is added to
omx video decoder proxy component to dump the raw frames from
ducati to mmc which can be analysed on PC tools for
debugging and quality assessment.
This feature supports enable/disable frame dump
at run time, the range can be specified.
The frames are dumped to data partition, and they can be
pulled over adb.
By default this feature is kept disabled to avoid security leaks.
How To Enable and Use#
(1) Uncomment the below 2 lines from Android.mk
And rebuild the omx proxy common component
(2) Before start playback, make sure that "data" folder has r/w
permissions. For this execute the below
mount -o rw,remount -t ext3 /dev/block/mmcblk0p1 /data/
chmod 777 /data/
(3) Set the property for number of frames to dump
eg: setprop debug.video.dumpframe 10:20
would dump frames from 10 to 20.
(4) Pull the frames to PC over adb
adb pull /data/frame_10.txt
(5) Analyse on PC tools.
Change-Id: I72c05f568daaaf88c36ceb79d158b65a2a02da1a
Signed-off-by: Sunita Nadampalli <sunitan@ti.com>
|
|/
|
|
|
|
|
|
|
|
|
|
| |
Description: VTC feature needs tunneling feature between
camera and encoder. This patch enables a9 side tunneling
infrastructure.
Index to enable VTC feature is appended to OMX_TI_Index.h
CQ: None
Change-Id: I57cb42d7269d498c93d17db8ac768788c2478af2
Signed-off-by: Phanish Hanagal <phanish.hs@ti.com>
|
|
|
|
|
|
|
|
|
|
| |
Trace was showing zero instead of the errorcode returned in
PROXY_paramCheck macro in omx_proxy_common.h
Trace was showing null instead of the checked condition in
RPC_paramCheck macro in omx_rpc_utils.h
Change-Id: Ib3e21ba5fdb12403bed38d3356c6b9293336b18f
Signed-off-by: Shivaraj Shetty <shettyshivaraj@ti.com>
|
|
|
|
|
|
|
| |
Adding a check for NULL buffer address in PROXY_UseBuffer()
and returning appropriate error on failing.
Signed-off-by: Saurabh Bipin Chandra <a0131926@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In case of ducati faults, all pending OMX calls should return
with OMX_ErrorHardware so that client is not blocked waiting
for any OMX call to return while ducati has faulted.
Also, all calls after the error hardware callbacks should
also get OMX_ErrorHardware.Just for GetState no error should
be returned, but the state returned should be OMX_StateInvalid.
Note# This has been ported from the below patch
https://android.dal.design.ti.com/#change,4852
Change-Id: I3efa845ea052bf4f4fb80b981f9e1701a7c87a31
Signed-off-by: Sunita Nadampalli <sunitan@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch adds support for passing multiple 2D virtual buffers by
passing a descriptor structure in pBuffer of OMX buffer header.
Interface only support is added for platform 1D and 2D buffers
which is envisioned to be used for passing ION handles
Signed-off-by: Abhishek Ranka <asranka@ti.com>
Conflicts:
omx_core/inc/OMX_TI_Index.h
Change-Id: I112fbcda623d89e5f20c2e8000caba99f76678ae
|
|
|
|
|
|
|
|
|
| |
Transfer pComponentUUID from Ducati in RPC_GetComponentVersion.
In this field is located information about ducati branch,
commit ID, commit message, build date and time
Signed-off-by: Ivan Evlogiev <x0099089@ti.com>
Change-Id: Ic0e0f8eb7330b265b62a6703b0321906dc54bf24
|
|
|
|
|
|
|
|
|
| |
In case of some error during GetHandle, there was a double
free of a pointer which was finally leading to a seg fault.
This patch rectifies that clean up path.
Change-Id: Ia24fc746d9a274ac0b7bea4b2dfbcde657b15c13
Signed-off-by: Sarthak Aggarwal <sarthak@ti.com>
|
|
|
|
|
| |
Change-Id: Ib2e253090258299ec21f9f17e1763be8bf3cc1dc
Signed-off-by: Iliyan Malchev <malchev@google.com>
|
|
|
|
|
|
|
| |
Patch moves all memmgr buffers to ion and maps them to userspace as and when required.
Change-Id: Iff8cc0759594918e5902625c052ea48d74bc07e1
Signed-Off-By: Sarthak Aggarwal<sarthak@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Set the alignment parameter for ion allocations to 4K. The
allocation size is already configured to be a multiple of
4K. This ensures that all the DOMX buffers can be mapped
into the user-space, and ensure that allocations are not
returned on a 4K boundary due to extraneous ion client
allocations from the same ion heap.
Change-Id: Ief40a6f30f48fccb315e073d28a2591d027e8243
Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Sarthak Aggarwal <sarthak@ti.com>
|
|
|
|
|
|
|
|
|
| |
The Stagefright queries DOMX in order to allocate buffers from gralloc with
correct usage flags.
Change-Id: I85dc7bd6bf6b3008a12470bb797df0cce8a370e0
Signed-off-by: Lakshman Gowda <lakshman79@ti.com>
Signed-off-by: Iliyan Malchev <malchev@google.com>
|
|
|
|
|
|
|
|
|
|
| |
Add support for android opaque color format as encoder input and
use color conversion routines to format to NV12 color format
which is passed on to remote OMX encoder component
Change-Id: I45be6a6ea3577cbe9a13e7df87bb399b16e5db29
Signed-off-by: Abhishek Ranka <asranka@ti.com>
Signed-off-by: Lakshman Gowda <lakshman79@ti.com>
|
|
|
|
|
| |
Change-Id: I3d0f40b68c2337e67271784c544db8979dcc0358
Signed-off-by: Sundar Raman <sunds@ti.com>
|
|
|
|
|
|
|
|
|
|
| |
request and allocate smaller-sized buffer in metadata mode recording.
the buffer is used to hold video_metadata_t structure (12 bytes). It is
a big saving from allocating an NV12 buffer size. Corresponding Ducati
side change on UseBuffer() is required.
Change-Id: I7e4730c0982e826def844e3cf5050d83063fa2a7
Signed-off-by: John Wei <a0272715@ti.com>
|
|
|
|
|
|
|
|
| |
Logs shared/observed on failures dont have proper error code.
UNDEFINED error code masks failure reason
Change-Id: I1a7ba901b1301e283f385db6117807501043bc9f
Signed-off-by: Devaraj Rangasamy <dev@ti.com>
|
|
|
|
|
| |
Change-Id: I1b6bb594740122c753d21eaa7a739a8e500ab57d
Signed-Off-By: Sarthak Aggarwal<sarthak@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
There was stale data in the metadata buffers during
initial few FBDs before omx component was ready to
fill the data. Seeing this in face detection usecase
where a few false rectangles are being shown when
the app is first opened.
Signed-off-by: Tyler Luu <tluu@ti.com>
Change-Id: I442d90a2e81ed0439ea0257850abfa9e0d0da1b1
Signed-off-by: Sundar Raman <sunds@ti.com>
|
|
|
|
|
| |
Change-Id: Iac07645546e352c37170eb46602bd1b6555fe85a
Signed-Off-By:Sarthak Aggarwal<sarthak@ti.com>
|
|
|
|
|
|
|
| |
The patch moves 1d bitstream buffers needed by video codecs from MemMngr to ION.
Change-Id: I56f010f904a64ae0d36bbc8925498f3748a0ab42
Signed-Off-By:Sarthak Aggarwal<sarthak@ti.com>
|
|
|
|
|
| |
Change-Id: Ic6d53f51f150de6c8c1324aae7464b984eb6ee52
Signed-Off-by:Sarthak Aggarwal<sarthak@ti.com>
|
|
|
|
|
| |
Change-Id: Ia63104950fdfbaf41254187dd0ee516fde8eb7d2
Signed-off-by: Iliyan Malchev <malchev@google.com>
|
|
Change-Id: I8f7a7eeece0e516efa486b77e9d97805c0e65d3e
Signed-off-by: Iliyan Malchev <malchev@google.com>
|