View source for Upgrading GStreamer version of NVIDIA Jetson

<seo title="Upgrading Jetson's GStreamer version | RidgeRun" titlemode="replace" metakeywords="GStreamer, Linux SDK, Linux BSP,  Embedded Linux, Device Drivers, Nvidia, Xilinx, TI, NXP, Freescale, Embedded Linux driver development, Linux Software development, Embedded Linux SDK, Embedded Linux Application development, GStreamer Multimedia Framework, GPU, GStreamer, NVIDIA, Jetson, TX1, TX2, Jetson Nano, Nano, NVIDIA Jetson Nano, Jetson AGX Xavier, Xavier, AI, Deep Learning, Jetson TX1, Jetson TX2, Jetson Xavier, NVIDIA Jetson Xavier, NVIDIA Jetson Orin, Jetson Orin, Orin, NVIDIA Orin, NVIDIA Jetson AGX Orin, Jetson AGX Orin, Jetpack 5.1.1, NVIDIA Jetson" metadescription="Steps to how to install a GStreamer version on the Jetson other than the one installed by the SDK."></seo>

{{NVIDIA Pref Partner logo and RR Contact}}

The following wiki guides on how to install a GStreamer version on the Jetson other than the one installed by the SDK. It also has information on how to build the Jetson's accelerated GStreamer elements so that they run correctly in the new version.

This has been tested on an NVIDIA Orin AGX using Jetpack 5.1.1. 
<br>
{{Ambox
|type=notice
|small=left
|issue='''Note: Although upgrading the device's GStreamer version can provide new features and/or remove bugs present in GStreamer, NVIDIA only officially supports their accelerated elements for the sdk-installable version.'''
|style=width:unset;
}}

== Upgrading GStreamer ==

More details at [[Setting_a_GStreamer_Alternative_Environment#GStreamer-1.16_through_GStreamer-1.19_non-invasive_method_-_gst-buil | Setting a GStreamer Alternative Environment-GStreamer-1.16 through GStreamer-1.19]]

By running these steps for the desired version, you'll have all the software elements built for the specific version. Do note that you might need to install some development libraries for a given element if it is not being built out of the box.

=== GStreamer 1.16-1.19 ===

<syntaxhighlight lang=bash>
git clone https://github.com/GStreamer/gst-build --branch 1.18.6
cd gst-build
meson builddir
ninja -C builddir
</syntaxhighlight>

=== GStreamer >= 1.20 ===

<syntaxhighlight lang=bash>
git clone https://gitlab.freedesktop.org/gstreamer/gstreamer.git -b 1.22
cd gstreamer
meson builddir
ninja -C builddir
</syntaxhighlight>

== Installing NVIDIA elements ==
* This guide only installs nvvidconv, the nvcompositor, and the nvdecoder/encoders, for others the process should be pretty similar, although you might need to patch the Makefile similarly to how this is done for some of these elements.
* Each element is independent of each other, you don't need to install all of them for one of them to work. However, the nvvidconv is used in most pipelines to convert from CPU memory to NVMM memory, so it is commonly needed for most use cases.

Go to https://developer.nvidia.com/embedded/jetson-linux-archive and select the corresponding L4T to the installed Jetpack. From there download the <code>Driver Package (BSP) Sources</code> tarball, and copy it over to the Jets

===Activate the environment===

* This is only needed if you didn't install the newer version and want to keep the SDK's version. At the end of this section, there are instructions on how to add each element to the development environment.

<syntaxhighlight lang=bash>
# This assumes that you only built the environment and didn't install it
cd gst-build 
# For >=1.20
cd gstreamer

ninja -C builddir devenv

</syntaxhighlight>

===Extract the sources===

<syntaxhighlight lang=bash>

cd ~ # Assuming that this is where you copied the public_sources.tbz2 tarball

tar xvf public_sources_3531.tbz2

cd Linux_for_Tegra/source/public/


</syntaxhighlight>

===NVencoder/decoder===

<syntaxhighlight lang=bash>
# NVdecoder/encoder
tar xvf gst-nvvideo4linux2_src.tbz2
cd gst-v4l2
make 
cd ..

# Add plugins to the path without installing
GST_PLUGIN_PATH=$GST_PLUGIN_PATH:$PWD/gst-v4l2/

</syntaxhighlight>

===Nvvidconv===

{{Ambox
|type=notice
|small=left
|issue= '''Note:''' This requires having DeepStream installed, if it isn't installed, use the sdk-manager to install it.
|style=width:unset;
}}<syntaxhighlight lang=bash>

tar xvf gst-nvvidconv_src.tbz2
cd gst-nvvidconv/

cat > nvvidconv_makefile.patch << EOF
--- Makefile
+++ Makefile
@@ -46,6 +46,8 @@
 INCLUDES += -I\$(CUDA_PATH)/include
 INCLUDES += -I\$(CUDA_PATH)/targets/\$(ARCH)-linux/include
 
+INCLUDES += -I/opt/nvidia/deepstream/deepstream/sources/includes/
+
 INCLUDES += -I./
 
 PKGS := gstreamer-1.0 \

EOF

patch < nvvidconv_makefile.patch
make 
cd ..

# Add plugins to the path without installing
GST_PLUGIN_PATH=$GST_PLUGIN_PATH:$PWD/gst-nvvidconv/

</syntaxhighlight>

===NvCompositor===

<syntaxhighlight lang=bash>

tar xvf gst-nvcompositor_src.tbz2
cd gst-nvcompositor/

cat > nvcompositor_makefile.patch << EOF
--- Makefile	2023-06-20 22:42:32.041529245 +0000
+++ Makefile	2023-06-20 22:42:01.429480422 +0000
@@ -35,7 +35,15 @@
 
 SRCS := $(wildcard *.c)
 
+ARCH := $(shell uname -m)
+ifdef CUDA_VER
+CUDA_PATH := /usr/local/cuda-$(CUDA_VER)
+else
+CUDA_PATH := /usr/local/cuda
+endif
+
 INCLUDES += -I./
+INCLUDES += -I/opt/nvidia/deepstream/deepstream/sources/includes/
 
 PKGS := gstreamer-1.0 \
 	gstreamer-base-1.0 \
@@ -54,7 +62,7 @@
 LDFLAGS = -Wl,--no-undefined -L$(LIB_INSTALL_DIR) -Wl,-rpath,$(LIB_INSTALL_DIR)
 LDFLAGS += -L$(CUDA_PATH)/targets/$(ARCH)-linux/lib/
 
-LIBS += `pkg-config --libs $(PKGS)`
+LIBS += `pkg-config --libs $(PKGS)` -lnvbufsurface -lnvbufsurftransform -lcudart
 
 all: $(SO_NAME)
 
EOF

patch < nvcompositor_makefile.patch
make 
cd ..

# Add plugins to the path without installing
GST_PLUGIN_PATH=$GST_PLUGIN_PATH:$PWD/gst-nvcompositor/

</syntaxhighlight>

{{ContactUs}}

[[Category:HowTo]][[Category:Jetson]][[Category:GStreamer]]