va: drop VAEntrypointMax enumeration.

This is an oversight, and that was supposed to be removed earlier.
User applications cannot rely on this value to statically allocate
an array of possible VA entrypoints. The correct interface is to
use vaMaxNumEntrypoints() and dynamically allocate that array.

Rationale: VA entrypoints values are maintained sparsely, so this
can yield to a large number overall.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

va: include <stdint.h> header.

Always include <stdint.h> header by default as we are using standard
fixed-size integral types from now on.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

vp8 dec: add clear notation for qindex

Signed-off-by: Zhong Li <zhong.li@intel.com>

vp8 dec: refine va_dec_vp8.h

add clear notation for: key_frame/VABoolCoderContextVPX.count/num_of_partitions/partition_size[0]
remove mb_skip_coeff since it can be deduced from mb_no_coeff_skip if it is required

Signed-off-by: Zhao, Halley <halley.zhao@intel.com>

libva: change the types of all fields in VABoolCoderContextVPX

BZ: 120268

Change the types of all fields in VABoolCoderContextVPX to
"unsigned char" to avoid potential bugs.

Change-Id: I483f115080cefe45ae49503ff5737c7dfff305a8
Signed-off-by: Tianmi Chen <tianmi.chen@intel.com>

Changes to VASliceParameterBufferVP8 to make it consistent with slice parameter buffer structure for other formats.

Added VASliceParameterBufferVP8 to describe data partition related fields.

Add VP8 decoding API

Added new profile, buffer types and data structures for VP8 decoding

Signed-off-by: Jonathan Bian <jonathan.bian@intel.com>
[Fixed conflicts and remove unused profiles]
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

do not request any GLX_DEPTH_SIZE in create_tfp_surface

Window depth was confused with OpenGL depth buffer.

Signed-off-by: Rinat <ibragimovrinat@mail.ru>

MPEG2ENC: Initialize the f_code based on supported level instead of hacked setting

The f_code of MPEG2 picture parameter indicates the max search range of the
motion vector prediction. The small value will cause the unoptimized motion
vector prediction. So remove the hack setting.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
(cherry picked from commit 4fa9b3ba9d45a021e282f761d061e1151158c3ea)

va: add new ABGR/XBGR fourcc codes.

Add VA_FOURCC_ABGR and its companion VA_FOURCC_XBGR definition. This
represents a packed RGB format with 32 bpp, i.e. 8-bit per component.
The components order in memory is A B G R.

Signed-off-by: Victor Ying <victor.ying@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

drm: improve check for authentication.

On Linux systems, the drmGetClient() function would return the thread ID
instead of the actual process ID.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

build: fix make dist.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

libva: bump version for development.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

libva 1.2.1

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

dummy_drv_video: add back VA_DRIVER_INIT_FUNC definition for __vaDriverInit_*()

Since commit ca1798e77b6fa99dbb010f202a057f75415afd52, dummy driver
was defining VA_DRIVER_INIT_FUNC symbol instead of __vaDriverInit_0_33
and therefore cannot be used for testing anymore. We restore the
definition of this symbol in configure.ac.

Signed-off-by: Vincent Bernat <bernat@luffy.cx>

tests: rename mpeg2enc to mpeg2vaenc

This fixes namespace issue with mpeg2enc from the mjpegtools.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66221
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

New version for development

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

libva 1.2.0

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

NEWS: updates

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

Remove the dummy driver dependency on X11

Otherwise when x11 is disabled, it can't be compiled correctly.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>

Remove the EGL dependency on X11

Otherwise when x11 is disabled, it can't be compiled correctly.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>

h264encode: fix mmap out of memmory issue

Change-Id: I9d7047d5ac66efefb31509f3ee52a9db28aada23
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: add more cmdline options

Change-Id: Ia0326c2649dcf6dfbce275539cb8f4221fa37501
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: fix ftell overflow issue when open large source YUV files

Change-Id: I30190d4e28bd643f00808a15846e50e8fca28764
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: coding 1920x1080 with crop

Change-Id: I4ba61943c9f2d246deb4bcb69a4781680f63b046
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: print correct number of RefPicList0/1

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: fix thread lockup issue

Change-Id: I6065525f76796603856f5cebf0468f4a59c4b5b5
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: fix thread lockup issue

Change-Id: I3e0f23b5301d8405297b4b88b27e9edccdab6063
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: fix idr_pic_id issue when encode multiple IDR frames

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: reset IDR poc to 0

Signed-off-by: Elaine Wang <elaine.wang@intel.com>

h264encode: dump reconstructed frame and calc PSNR

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: use frame height not the height from VAImage which may be wrong

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: correct negative POC after a new IDR

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: fix encode corruption issue with multiple IDR (-idr_period XX)

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

h264encode: add some comments

Signed-off-by: Elaine Wang <elaine.wang@intel.com>

h264encode: refine calculation logic of reference frame setting

Signed-off-by: Elaine Wang <elaine.wang@intel.com>

h264encode: fix wrong print information

Signed-off-by: Elaine Wang <elaine.wang@intel.com>

AVCENC: Pass the H264 RefPicList0/1 by using slice_param instead of hacked DPB reference

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
(cherry picked from commit 968a9bf006679ad8d752fa1ddcd4bb80fc8c5e51)

Fix possible memory leak

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit 0b07495fc24646b47952b8df7d4e6d2df2a0f72c)

va.c: need to check the symbol before calling it

This is to fix Klockwork critical issue

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 40e5e45edf97049137bca014575607b52d3bca78)

va: fix vaErrorStr() to support new error codes.
(cherry picked from commit 7b5c35273dc0e8e0e86cf40cf2a398119a766397)

va_x11: fix klockscan issue (details is bellow)

File:        va_x11.c
Function:    va_DisplayContextGetDriverName -> va_DRI2GetDriverName -> dri2_util.c:isDRI2Connected
Line:        154 -> 84 -> dri2_util.c:184
Code:
TRACEBACK
va_x11.c:151: driver_name is NULL
va_x11.c:154: 'driver_name' is passed to function 'va_DRI2GetDriverName'.
dri2_util.c:84: 'driver_name' is passed to function 'isDRI2Connected'.
dri2_util.c:184: 'driver_name' is explicitly dereferenced.

Signed-off-by: Austin Yuan <shengquan.yuan@gmail.com>
(cherry picked from commit c87699f32c6c79905a2fd8675cd19f22e1f4b024)

vatrace: print target_percentage

Change-Id: I220b2fa7ad6907ab0ebda3671baeeab7fa36cc00
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 82e7dc8280da22c88a28f0a4d429f28fd99f9ced)

vatrace: trace coded buffer create/destroy

Change-Id: If0388cd5e2e141184d88ff5de5612d15e51e3d20
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 880c916092c6d68096c95e49405714d8a21f9322)

vatrace: dump unknonw private misc parameters

Change-Id: I33bbed3a1d3f297297687197f623d6c08ee8eba9
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit df5bfb6c6f248c0bcf6f2f9a74e9ffba4b0acf15)

vatrace: remove the delimiter of every elements dump

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 132f22ae9b42b4657361c87b5c7d5b14a26a52a9)

vatrace: refine the format of buf data dump

Signed-off-by: Elaine Wang <elaine.wang@intel.com>
(cherry picked from commit 2292299fe00036eb6055a84846f63f36cbae27d9)

vatrace: refine VAEncPackedHeaderParameterBufferType format

Change-Id: I1794a9c09a88b61d7abce970d7cb9a8f101162c6
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 4853e4a3a6cc5218e032a19d729096862db82bc6)

vatrace: print VAEncPackedHeaderParameterBufferType

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 82ffa4b5bfb6cf59e90eedadbdfe3efdaced35e0)

libva: Add VABuffer dumping for JPEG parameters

BZ: 98966

Dump VABuffer of JPEG parameters in libva

Change-Id: I87450d92db402a4730e0c0433215c857bc5b494d
Signed-off-by: Cheng Yao <yao.cheng@intel.com>
(cherry picked from commit 40ffef8d1fcbbc709472dc1342d7e3e983ae5d96)

vatrace: only print info of valid reference frames

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 67c489f323188df1370f8ba2d4063bc86153cce7)

vatrace: dump flags of VAPictureH264

Signed-off-by: Elaine Wang <elaine.wang@intel.com>
(cherry picked from commit e1d86f000749d5fdbd6e36b55260d3fae096e457)

LIBVA_DRIVER_NAME still needs va_getDriverName to open DRM device

va_getDriverName is bypassed when LIBVA_DRIVER_NAME is set, but driver
needs DRM device fd which is only opened in va_getDriverName

Change-Id: I4fe11edf9d8c328806255d4c93313cf99655a91d
Signed-off-by: Fei Jiang <fei.jiang@intel.com>
(cherry picked from commit 04706eb13ae410b53b6b38150413d544b7fdb524)

va_trace: refine the log format

Signed-off-by: Elaine Wang <elaine.wang@intel.com>
(cherry picked from commit 15a0fca64456d8c7de1f7aca2cb747c4929690cd)

vatrace: print function name for vaMapBuffer trace with coded buffer

also fall into default va_TraceVABuffers if the buffer type falls into
"default"

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit acbf7d0c4830763e7c3be0e71c9da972f1aea7a5)

vatrace: buffer_type_to_string: add missing VABufferType

sync with VABufferType in va/va.h.

Signed-off-by: Adrian Marius Negreanu <adrian.m.negreanu@intel.com>
(cherry picked from commit d5160173f9dd5a37c793518d1d483b09d278d90c)
[Remove unsupported types on master]
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

va_parseConfig: still need to check env_value

env_value may be NULL which means the caller doens't
care the value of the environment variable, just check
if it is set.

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 2e11d2273b2974a7d1959cbcaf8db5b8e9aedd9e)

vatrace: fix build issue

Change-Id: I1727528df6aaf48209774bada5cf136cfa1e4ee3
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 1e0421f5ce01790e700940310aafb05e4c18d8f0)

va_trace: dump basic_unit_size/rc_flags in RateControl data struct

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 4a78cb4e4a2d0e5c9231c7f532d2a733cf3bfce3)

vatrace: Fix encoded coded buffer dump function

libva trace can't dump coded buffer. It's caused by the wrong
decison in vaTraceEndPicture function which only dump coded buffer
when surface dumping is enabled.
Signed-off-by: Elaine Wang <elaine.wang@intel.com>
Change-Id: I8b0ec01e59faff3bfd5345d95cbf0e781f6f3706
(cherry picked from commit 202385430f29b3d3d356d86888a0f3979986b4e7)

vatrace: fix vatrace crash during video recording

Video recording crash when va trace is enabled. It's caused by
using wrong buffer type to trace slice buffer. This patch distinguish buffer
type VAEncSliceParameterBuffer from VAEncSliceParameterBufferH264 by
the buffer size.
Signed-off-by: Elaine Wang <elaine.wang@intel.com>
Change-Id: Ibc5c6a17b155e305151a25580b4e613702665ad1
(cherry picked from commit b5b1933f60dfec8cf2cff19082362accbe6fce1d)

va_trace: allow surface trace to dump part of surface content

To debug/isolate video record quality issue (blocky/corruption, etc),
we always need to dump the YUV image from camera. LibVA has such
functionality, but due to storage bandwidth issue (~10M Byte/second),
it is hard to write 1080P 30fps YUV data to disk (~88M Byte/second).
Now refine the functionality to only dump part of the image, and
won't need to modify the code in the default build.

The method is:
edit /etc/libva.conf to include bellow (all export them):
LIBVA_TRACE_SURFACE=/data/enc-XXX.yuv
LIBVA_TRACE_GEOMETRY=320x240+20+20

It will only save 320x240 from (20,20) of the source surface to file
/data/enc-XXX.yuv.

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 5190a7465756a41808a562c8eccec521dbf77d8c)

va_trace_surface: dumpe surface content after vaEndPicture call

vaEndPicture will kick HW decode/encode, and at this time use CPU
to dump surface content can keep GPU/CPU in parallel.

Also for encode trace, don't need to wait current frame completions
but for decode, we must wait decode till it is completed

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>
(cherry picked from commit 0af365de4c03c12fc20183093a944f7513a1c7be)

Fix va_fool.c klocwork issue

BZ: 83882

Fix klocwork issue that complaine array 'fool_context[idx].fool_buf'
of size 43 may use index value(s) 0..255

Change-Id: I196a36c0067e2b7e390567a203ef1520af5b873f
Signed-off-by: Elaine Wang <elaine.wang@intel.com>

klockwork: Fix libva klocwork scaning issues

Fix NULL pointer dereference risk issues and uninitialized variables

Signed-off-by: Elaine Wang <elaine.wang@intel.com>
Change-Id: I839a9def7b08c41c1f075e4cd1cc06e8bb0c8c9e

tests: cope with new vaQuerySurfaceAttributes() API.

Also factor the code to check whether a specific image format or
surface attribute exists. Cache the list of supported image formats
and surface attributes.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

test: add color conversion test option in putsurface

- command line:  ./putsurface --fmt1 format1 --fmt2 format2
              or ./putsurface -1 format1 -2 format2
- example:       ./putsurface -1 NV12 -2 YV12
- implementation:
  - vaSurface is created with fmt1
  - get image (with fmt2) from vaSurface,
    it leads to color conversion
  - put the above image to a temp surface
    color conversion may happen
  - render the above temp surface instead of original one

tests: cope with new vaCreateSurfaces() API.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Signed-off-by: Haitao Huang <haitao.huang@intel.com>
Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

Add default implementation for vaQuerySurfaceAttributes().

Add an initial implementation for vaQuerySurfaceAttributes() in terms
of vaGetSurfaceAttributes(). This is an inefficient approach because
it dynamically allocates the whole set of surface attributes we guessed
are available.

This code is to be removed once the VA drivers are updated.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: add vaQuerySurfaceAttributes().

vaQuerySurfaceAttributes() is the recommended way to query the driver for
all surface attributes that are supported by the underlying hardware. This
is more useful than guessing the attribute values and testing them through
vaGetSurfaceAttributes().

The older vaGetSurfaceAttributes() function is hereby removed. However, the
relevant VA driver hook is kept for now, and will be removed/replaced for a
future update.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Added new VA_RT_FORMAT_RGBP and VA_FOURCC_411R

SurfaceAttribMemoryType: add ION memory type for Android

ION is the new Android memory manager for cross-components buffer
sharing. Adding this memory type for it.

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

ASurfaceAttribExternalBuffers flags: add more memory attribute flags

The external user pointer buffers may be allocated as cacheable or
non-cacheable, with the addtitional cache attribute flags, it can help
driver implementation to set correct cache attribute into device
memory controller (e.g. MMU)

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

Changed values for VASurfaceAttribMemoryType from fourcc to bit fields.

Added new RT formats and fourcc codes for pixel formats.

Added new surface attributes to enable VA surface creation from external buffers.

Add RGB RT format definition for video process

RGB format surface is required for color
space conversion from YUV to RGB.

signed-off-by: Li,Xiaowei <xiaowei.a.li@intel.com>

tpi: clean the old interface, and use new one

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

Add new fourcc codes

Added fourcc codes for YV24 and YV32 (4:4:4 planar formats)

Signed-off-by: Jonathan Bian <jonathan.bian@intel.com>

va_TraceCreateSurface->va_TraceCreateSurfaces: correct name change

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

API: add compat glue for older vaCreateSurfaces().

API: add surface attributes.

test: update MPEG-2 encoding case

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

tests: disable v4l demo for now.

vaPutSurface test on Android: fix build isste on Android JB

Change-Id: Ibfb3d6a653f9af924422edb052e751b84015ffd0
Signed-off-by: Austin Yuan <shengquan.yuan@gmail.com>

h264encode: fix build warning on Android

Change-Id: Iee2886bde2362e75d7e1837211553d9672bba650
Signed-off-by: Austin Yuan <shengquan.yuan@gmail.com>

h264encode: remove the leftover file h264encode_android.cpp

Signed-off-by: Austin Yuan <shengquan.yuan@gmail.com>

h264encode: fix some issues

1. reference frame setting looks odd
2. fix the issue that the first 16 frames are encoded twice

Signed-off-by: Elaine Wang <elaine.wang@intel.com>

test/h264encode: refine the h264encode to support advanced mode

Signed-off-by: Austin Yuan <shengquan.yuan@gmail.com>

test/putsurface_x11.c: don't hard code the DISPLAY to ":0.0"

Signed-off-by: Austin Yuan <shengquan.yuan@gmail.com>

mpeg2enc: find the proper profile and level

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

mpeg2enc: Fix the picture header for B frame

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

mpeg2enc: Fix time_code_seconds in time_code

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

mpeg2enc: temporal_referenc and GOP header follow MPEG-2 Spec.

temporal_reference is reset to zero at each group of pictures header

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

mpeg2enc: Silence compiler warning

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

test/encode: Add -Wall option to default build flags

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

mpeg2enc: Simplify the mapping between display order and encoding order

mode 0:
D: IIIII...
E: IIIII...

mode 1:
D: IPPPP...
E: IPPPP...

mode 2:
D: IBBPBBPBBPBBPBBPIBBP...
E: IPBBPBBPBBPBBPBBIPBB...

B frame is supported now

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

mpeg2enc: set the frames between I and P frames

The default valude is 2

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

mpeg2enc: forward and backword referenc follow the picture type

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>