Commit graph

781 commits

Author SHA1 Message Date
George Kiagiadakis
fbe0d07068 gstomx: remove gst_omx_buffer_set_omx_buf/get_omx_buf
They are no longer used anywhere
2019-04-25 09:09:42 +05:30
George Kiagiadakis
783e58fc48 omxbufferpool: refactor to allow memory sharing
One big restriction of the OMX buffer pool has always been
that the GstMemory objects were flagged with NO_SHARE.
This was because the buffer pool needed to be sure that when
a buffer returned to the pool, it would be safe to release the
OMX buffer back to OpenMAX.

With this change, this is no longer a restriction. What this
commit introduces is a new allocator that allows us to track
the GstMemory objects independently. Now, when a buffer returns
to the pool, it is not necessary for the memory to be released
as well. We simply track the memory's ref count in the allocator
and we return the OMX buffer back when the memory's ref count
drops to 0.

The reason for doing this is to allow implementing zero-copy
transfers in situations where we may need to copy or map a
certain region of the buffer. For instance, omxh264enc ! h264parse
should be possible to be zero-copy by using an OMX buffer pool
between them.
2019-04-25 09:09:40 +05:30
Guillaume Desmottes
3018ea5843 omxbufferpool: fix memory mapping with offset
gst_memory_map() is already adding the offset to the mapped pointer.
Doing it in the memory implementation was resulting in the offset being
accounted twice.

It doesn't matter yet as we are only creating memory without offset for
now but it will once we'll start sharing OMX memories.
2019-04-23 15:32:05 +05:30
Julien Isorce
18927f33ee Fixes build with omx >= 1.2.0
gstomx.c:1405:10: error: ‘OMX_IndexParamCustomContentPipe’ undeclared (first use in this function)
    case OMX_IndexParamCustomContentPipe

Some enums have been deprecated in 1.2.0

https://gitlab.freedesktop.org/gstreamer/gst-omx/issues/27
2019-04-16 12:26:03 -07:00
Tim-Philipp Müller
6ce69859c5 omx: fix autotools build for generic target
gstomxvideoenc.c:2874:7: error: "USE_OMX_TARGET_ZYNQ_USCALE_PLUS" is not defined, evaluates to 0 [-Werror=undef]
 #elif USE_OMX_TARGET_ZYNQ_USCALE_PLUS

Works on meson because it doesn't use -Wundef
2019-04-10 00:20:37 +01:00
Guillaume Desmottes
7d1b77b0c4 omx: disable OMX_API_TRACE code if gst debug is disabled
No need to create debug structs which won't be used as DEBUG macros are
no-op.
2019-03-26 15:15:21 +01:00
Guillaume Desmottes
01bf661d6e omx: log Get/SetParameter/Config calls
Extend OMX_API_TRACE by logging component configuration calls.
2019-03-26 15:15:21 +01:00
Guillaume Desmottes
b53c91d9ef omxbufferpool: don't use CAT_PERFORMANCE if pool will copy
This was the single place where this category was used in gst-omx so
most users, including me, are generally not turning it and were missing this
important information from logs.

The copying code uses gst_video_frame_copy() which is already logging
with CAT_PERFORMANCE so we can still have this information when using
only this debug category.
2019-03-26 15:15:21 +01:00
Guillaume Desmottes
6feac3ed26 omxvideodec: add debug if proposed pool can't provide enough buffers
We were silently ignoring the pool which was pretty confusing when
debugging.
2019-03-26 15:15:21 +01:00
Charlie Turner
7412b39765 omx: Add hardware classifiers to encoders/decoders 2019-03-26 12:17:26 +00:00
Guillaume Desmottes
9ff4dc533f omxvideo: use GST_VIDEO_CAPS_MAKE() for template caps
Simplify the code and so we advertise the formats actually supported by
gst-omx.
2019-01-25 23:48:11 +00:00
Guillaume Desmottes
09bc1d630f omxvideoenc: validate cpb-size and initial-delay
cpb-size cannot be smaller than initial-delay.
2019-01-25 11:48:44 +01:00
Nicolas Dufresne
aff131dcee omxvideodec: Remove duplicated QoS code
The 'finish' function do the exact same check / drop, there is no
need to duplicate this here.
2019-01-25 00:50:06 +00:00
Nicolas Dufresne
4159ca2bc5 omxvideodec: Remove dead code
The omxvideodec base class have a totally unused prepare_frame() vritual
function, remove it.
2019-01-25 00:50:06 +00:00
Varunkumar Allagadapa
150e7271ab omxvideoenc: add adaptive gop-mode option
Added adaptive gop-mode option to ZYNQ_USCALE_PLUS encoder properties
2019-01-25 00:49:22 +00:00
Nicolas Dufresne
4f6d788999 omxvideoenc: Add dynamic framerate support
Instead of going through a full reset, try and change the framerate
config on the encoder when only the framerate have change.
2019-01-25 00:47:45 +00:00
Varunkumar Allagadapa
09ad21b26c omxvideoenc: Add dynamic IDR insertion support on zynq
As the pi, the zynq has its own API to request keyframe.
2019-01-09 19:37:36 +00:00
Guillaume Desmottes
9dc359d0c1 omxbufferpool: fix race when releasing input buffers
If buffers were released from the pool while
gst_omx_video_enc_handle_frame() was waiting for new buffers,
gst_omx_port_acquire_buffer() was never awaken as the buffers weren't
released through OMX's messaging system.

GQueue isn't thread safe so also protect it with the lock mutex.
2019-01-08 16:04:31 +01:00
Guillaume Desmottes
3828d9769c omxbufferpool: fix early input buffer release
We used to track the 'allocating' status on the pool. It is used while
allocating so output buffers aren't passed right away to OMX and input
ones are not re-added to the pending queue.

This was causing a bug when exporting buffers to v4l2src. On start
v4l2src acquires a buffer, read its stride and release it right away.
As no buffer was received by the encoder element at this point, 'allocating'
was still on TRUE and so the the buffer wasn't put back to the pending
queue and, as result, no longer available to the pool.

Fix this by checking the active status of the pool instead of manually
tracking it down. The pool is considered as active at the very end of
the activation process so we're good when buffers are released during
the activation.
2018-12-21 17:16:31 +01:00
Guillaume Desmottes
839fbfebab omx: fix OMX_EventBufferFlag OMX_API_TRACE struct
The GType was missing from the second field of the struct.
2018-11-26 10:41:53 +01:00
Guillaume Desmottes
6207459e0e omx: rename OMX_PERFORMANCE debug cat to OMX_API_TRACE
This debug category can now be used to track more OMX calls and events
so best to rename it to something more generic.

https://bugzilla.gnome.org/show_bug.cgi?id=797171
2018-09-26 09:20:41 +02:00
Guillaume Desmottes
3a16094ab3 omx: log OMX commands with OMX_PERFORMANCE debug category
It has been useful to have a clear raw and structured view of the gst
<-> OMX exchanges when debugging.

https://bugzilla.gnome.org/show_bug.cgi?id=797171
2018-09-19 10:52:25 +02:00
Guillaume Desmottes
68f18341f6 omx: factor out gst_omx_component_send_command()
No semantic change. I'm going to add extra debug in this function.

https://bugzilla.gnome.org/show_bug.cgi?id=797171
2018-09-19 10:52:25 +02:00
Guillaume Desmottes
9b1598e5ec omx: log OMX events with OMX_PERFORMANCE debug category
It has been useful to have a clear raw and structured view of the gst
<-> OMX exchanges when debugging.

https://bugzilla.gnome.org/show_bug.cgi?id=797171
2018-09-19 10:52:25 +02:00
Guillaume Desmottes
313deb4ecf omx: rename log_omx_performance() to log_omx_performance_buffer()
I'm about to log more things under this category

https://bugzilla.gnome.org/show_bug.cgi?id=797171
2018-09-19 10:52:25 +02:00
Nicolas Dufresne
a84badc5c4 omxvideoenc: Remove spurious locking
The method we call in the context of pushing a buffer are all thread
safe. Holding a lock would prevent input buffers from being queued while
pushing.

https://bugzilla.gnome.org/show_bug.cgi?id=715192
2018-09-10 17:09:11 -04:00
Nicolas Dufresne
d80504afcc omxvideoenc: Remove unneeded size check
We only enter this branch if nFilledLen > 0, there is not need
to check again.

https://bugzilla.gnome.org/show_bug.cgi?id=715192
2018-09-10 17:09:11 -04:00
Nicolas Dufresne
35abdd16d1 omxvideodec: Remove spurious unlock in error case
This was forgotton in previous patch. We no long hold the lock when goto
invalid_buffer is called.

https://bugzilla.gnome.org/show_bug.cgi?id=715192
2018-09-10 17:09:11 -04:00
Nicolas Dufresne
04e1f76b60 omxvideodec: don't hold the stream lock when trying to push a frame
The base class methods will lock this properly when needed, there seems
to be no need to lock it explicitly.

This allows the patch in gstvideodec for unlocking the stream lock
when pushing buffers out to work.

https://bugzilla.gnome.org/show_bug.cgi?id=715192
2018-08-31 17:42:07 -04:00
Guillaume Desmottes
f108eeafde omxvideodec: don't import OMX buffers from downstream
We already have code configuring the encoder stride and slice height
when receiving the first buffer from upstream.
We don't have an equivalent when the encoder is exporting its buffers to the
decoder.

There is no point adding it and making the code even more
complex as we wouldn't gain anything by exporting from the encoder to
the decoder. The dynamic buffer mode already ensures 0-copy between OMX
components.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
7be54ad091 omxvideoenc: implement dmabuf export on input buffers
Propose pool upstream so input buffers can be allocated by the port and
exported as dmabuf.

The actual OMX buffers are allocated when the pool is activated, so we
don't end up doing useless allocations if the pool isn't used.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
34bc02e397 omx: allow gst_omx_port_acquire_buffer() to not wait for buffers
Will be needed to implement GST_BUFFER_POOL_ACQUIRE_FLAG_DONTWAIT.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
c89b54fe78 omxvideodec: don't import non-dmabuf when dec is in dmabuf mode
Fix 'omxh264dec ! videocrop' pipeline.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
be5ec66a84 omxvideodec: factor out gst_omx_try_importing_buffer()
No semantic change, just make the code clearer and improve debug output.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
7d8a314dff omxvideodec: fix gst_video_info_from_caps() caps assertion
The "use buffers" code path uses gst_video_info_from_caps() which is
asserting if caps is NULL (because pool was rejected).

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
9ed1c2e75e omxvideodec: fix pool caps reference stealing
gst_buffer_pool_config_get_params() doesn't ref the returning caps;
so gst_caps_replace() was unreffing the reference owned by the pool.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
0996019c07 omxvideodec: prevent timeout when shutting down because of pending out buffers
The OMX transition state to Loaded won't be complete until all buffers
have been freed. There is no point waiting, and timeout, if we know that
output buffers haven't been freed yet.

The typical scenario is output buffers being still used downstream
and being freed later when released back to the pool.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
86a6703d91 omxbufferpool: reference the OMX component
Now that the pool is responsible of freeing the OMX buffers, we need to
ensure that the OMX component stay alive while the pool is as we rely on
the component to free the buffers.
The GstOMXPort is owned by the component so no need to ref this one.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
8efa0951ff turn GstOMXComponent to a GstMiniObject
Will use it for refcounting.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
7b6be347a6 omxbufferpool: deallocate OMX buffers when stopping
The pool is stopped when all the buffers have been released. Deallocate
when stopping so we are sure that the buffers aren't still used by
another element.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
d2f7e21ab9 omx: call gst_omx_buffer_unmap() when handling BUFFER_DONE
When using a input buffer pool, the buffer may be released to the pool when
gst_omx_buffer_unmap() is called. We need to have buf->used unset at
this point as the pool may use it to check the status of the pool.

{Empty,Fill}BufferDone is called from OMX internal threads while
messages are handled from gst elements' thread. Best to do all this
when handling the message so we don't mess with OMX threads and keep
the original thread/logic split.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
3b92b22554 omxvideo{enc,dec}: stop calling shutdown() in change_state
This is no longer needed since we implemented close() vfuncs as the
encoder/decoder base class already take care of calling close() (which
is calling shutdown()) in its own change_state implementation.

We also move the shut down of the component from PAUSED_TO_READY to READY_TO_NULL.
By doing so upstream will have already deactivated the pool from the
encoder and so won't be preventing the OMX state change as the buffers
will all be released.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
457e1b9a5e omx: factor out gst_omx_buffer_get/set_omx_buf()
Move the qdata code to helper functions as I'm going to need them in
omxvideoenc to implement dmabuf export.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
092788a968 omxvideoenc: factor out gst_omx_video_enc_set_to_idle()
No semantic change. We'll have to use this when the input pool is
activated so we can allocate buffers.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
7d3b0cd1e8 omxvideoenc: factor out gst_omx_video_enc_deallocate_in_buffers()
Will add extra code when adding input buffer pool.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
998bfbb906 omx: add pBuffer to OMX_PERFORMANCE logs
Can be useful to check the fd being passed when using dmabuf.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
f0964dfbdb omx: factor out gst_omx_port_set_dmabuf()
No semantic change. I also made the debug message a bit clearer.

https://bugzilla.gnome.org/show_bug.cgi?id=796918
2018-08-30 10:59:30 +02:00
Guillaume Desmottes
40ae47df5e omx: wait for flush complete and buffers being released when flushing
When flusing we should wait for OMX to send the flush command complete event
AND all ports being released.
We were stopping as soon as one of those condition was met.

Fix a race between FillThisBufferDone/EmptyBufferDone and the flush
EventCmdComplete messages. The OMX implementation is supposed to release
its buffers before posting the EventCmdComplete event but the ordering
isn't guaranteed as the FillThisBufferDone/EmptyBufferDone and
EventHandler callbacks can be called from different threads (cf 2.7
'Thread Safety' in the spec).

Only wait for buffers currently used by OMX as some buffers may not be
in the pending queue because they are held downstream.

https://bugzilla.gnome.org/show_bug.cgi?id=789475
2018-08-29 12:18:39 -04:00
Guillaume Desmottes
3d55051da3 omx: factor out should_wait_until_flushed()
No semantic change. Makes the code easier to understand and I'm about to
change the waiting condition.

https://bugzilla.gnome.org/show_bug.cgi?id=789475
2018-08-29 12:18:39 -04:00
Guillaume Desmottes
626a481e01 omxvideoenc: pause component when flushing
As stated in the spec ("6.1.3 Seek Event Sequence") we should pause
before flushing.
We were pausing the decoder but not the encoder so I just aligned the
two code paths.

https://bugzilla.gnome.org/show_bug.cgi?id=797038
2018-08-28 15:52:14 +02:00