Build

Compiling the Kernel (Kernel 6.1)

This topic contains instructions for compiling the Linux kernel source in your Linux SDK product.

1. Set up the environment macros: The LOCAL_VERSION environment variable appends -rt-tegra to the kernel version and modules.

   export KERNEL_NAME="nvidia-6.1"
   export ARCH=arm64
   export CROSS_COMPILE="${PWD}/toolchains/aarch64--glibc--bleeding-edge-2024.02-1/bin/aarch64-buildroot-linux-gnu-"
   export LOCALVERSION="-rt-tegra"
   export n_threads=`nproc`
   export TEMP_DIR=/tmp/$KERNEL_NAME
   export NV_BUILD_SYSTEM_TYPE="embedded-linux"
   

2. Prepare the output folders.

   cd ${NV_WORKSPACE}/drive-linux/
   rm -rf $TEMP_DIR/kernel
   mkdir -p $TEMP_DIR/kernel/$KERNEL_NAME $TEMP_DIR/kernel/conftest
   cp -rf kernel/nvidia-6.1/source/* $TEMP_DIR/kernel/$KERNEL_NAME
   cp -rf kernel/common/nvidia-oot $TEMP_DIR/kernel
   cp -rf kernel/common/nvgpu $TEMP_DIR/kernel
   cp -rf kernel/common/hwpm $TEMP_DIR/kernel
   cp -rf kernel/common/nvidia-oot/scripts/conftest $TEMP_DIR/kernel/conftest/nvidia
   

3. Apply the rt-patches and SoC-specific patches.

   cd $TEMP_DIR/kernel/$KERNEL_NAME
   bash scripts/generic-rt-patch.sh apply-patches
   

4. Install the packages to prepare the kernel building workspace.

   apt-get -f -y install imagemagick graphviz dvipng python3-venv fonts-noto-cjk latexmk librsvg2-bin texlive-xetex flex bison kmod
   

5. Configure and build the kernel.

   For dev_nsr PCT:

   export NV_BUILD_KERNEL_CONFIG_NAME=defconfig
   

   For prod_nsr and test_nsr PCTs:

   export NV_BUILD_KERNEL_CONFIG_NAME=tegra_prod_defconfig
   

   Build the kernel.

   cd $TEMP_DIR/kernel/$KERNEL_NAME
   make ARCH=arm64 CROSS_COMPILE=${CROSS_COMPILE} O=$TEMP_DIR/kernel/out-linux $NV_BUILD_KERNEL_CONFIG_NAME
   make ARCH=arm64 CROSS_COMPILE=${CROSS_COMPILE} O=$TEMP_DIR/kernel/out-linux -j$n_threads
   

6. Build OOT modules.

   cd $TEMP_DIR/kernel
   export IGNORE_MISSING_MODULE_SYMVERS=1
   export NV_BUILD_CONFIGURATION_EXPOSING_T26X=1
   
   make src=$TEMP_DIR/kernel/conftest/nvidia obj=$TEMP_DIR/kernel/conftest/nvidia CC=${CROSS_COMPILE}gcc LD=${CROSS_COMPILE}ld NV_KERNEL_SOURCES=$TEMP_DIR/kernel/$KERNEL_NAME NV_KERNEL_OUTPUT=$TEMP_DIR/kernel/out-linux -f $TEMP_DIR/kernel/conftest/nvidia/Makefile
   
   ARCH=arm64 CC=${CROSS_COMPILE}gcc LD=${CROSS_COMPILE}ld make -j$n_threads -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/hwpm/drivers/tegra/hwpm srctree.hwpm=$TEMP_DIR/kernel/hwpm kernel_name=oot system_type=$NV_BUILD_SYSTEM_TYPE V=1 modules CONFIG_TEGRA_OOT_MODULE=m CONFIG_TEGRA_VIRTUALIZATION=y srctree.nvconftest=$TEMP_DIR/kernel/conftest
   
   ARCH=arm64 CC=${CROSS_COMPILE}gcc LD=${CROSS_COMPILE}ld make -j$n_threads -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/nvidia-oot srctree.nvidia-oot=$TEMP_DIR/kernel/nvidia-oot srctree.nvidia=$TEMP_DIR/kernel/nvidia-oot srctree.hwpm=$TEMP_DIR/kernel/hwpm srctree.nvconftest=$TEMP_DIR/kernel/conftest kernel_name=oot system_type=$NV_BUILD_SYSTEM_TYPE KBUILD_EXTRA_SYMBOLS=$TEMP_DIR/kernel/hwpm/drivers/tegra/hwpm/Module.symvers V=1 modules CONFIG_TEGRA_OOT_MODULE=m CONFIG_TEGRA_VIRTUALIZATION=y
   
   ARCH=arm64 CC=${CROSS_COMPILE}gcc LD=${CROSS_COMPILE}ld make -j$n_threads -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/nvgpu/drivers/gpu/nvgpu srctree.nvidia-oot=$TEMP_DIR/kernel/nvidia-oot srctree.nvidia=$TEMP_DIR/kernel/nvidia-oot srctree.nvconftest=$TEMP_DIR/kernel/conftest kernel_name=oot system_type=$NV_BUILD_SYSTEM_TYPE KBUILD_EXTRA_SYMBOLS=$TEMP_DIR/kernel/nvidia-oot/Module.symvers V=1 modules CONFIG_TEGRA_OOT_MODULE=m CONFIG_TEGRA_VIRTUALIZATION=y
   

7. Consolidate the built kernel modules.

   export INSTALL_MOD_PATH=$TEMP_DIR/kernel/out-linux
   make ARCH=arm64 -C $TEMP_DIR/kernel/out-linux O=$TEMP_DIR/kernel/out-linux INSTALL_MOD_STRIP=1 modules_install
   make ARCH=arm64 -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/nvidia-oot INSTALL_MOD_STRIP=1 modules_install
   make ARCH=arm64 -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/hwpm/drivers/tegra/hwpm INSTALL_MOD_STRIP=1 modules_install
   make ARCH=arm64 -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/nvgpu/drivers/gpu/nvgpu INSTALL_MOD_STRIP=1 modules_install
   

8. Recompile the display kernel modules.

   Display provides two different variants of drivers in DriveOS.

   • prod_nsr and test_nsr use the tegradisp driver (tegradisp.ko).

   • dev_nsr uses OpenRM (nvidia.ko, nvidia_modeset.ko, nvidia_drm.ko).

   1. For tegradisp (prod_nsr/ test_nsr):

   cd $TEMP_DIR/kernel
   tar -xf $NV_WORKSPACE/drive-linux_src/tegradisp-sources.tar.gz
   cd tegradisp-opensrc
   
   ARCH=arm64 CC=${CROSS_COMPILE}gcc \
   LD=${CROSS_COMPILE}ld make -j$n_threads \
   -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/tegradisp-opensrc \
   srctree.nvidia=$TEMP_DIR/kernel/nvidia-oot kernel_name=oot \
   srctree.nvconftest=$TEMP_DIR/kernel/conftest \
   KBUILD_EXTRA_SYMBOLS=$TEMP_DIR/kernel/nvidia-oot/Module.symvers \
   V=1 modules CONFIG_TEGRA_OOT_MODULE=m CONFIG_TEGRA_VIRTUALIZATION=y
   
   # Install step
   make ARCH=arm64 -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/tegradisp-opensrc INSTALL_MOD_STRIP=1 modules_install
   

   2. For OpenRM (dev_nsr):

   # Unify headers and Module.symvers for display module compilation:
   cd $TEMP_DIR/kernel
   cat nvidia-oot/Module.symvers >> out-linux/Module.symvers
   rsync -avzpq nvidia-oot/include/ out-linux/include
   
   # Untar the display source code.
   # The NVIDIA-kernel-module-source-<Version>.tar.xz source code is supplied as a .xz file. Untar the file. Its location is ${NV_WORKSPACE}/drive-linux_src/.
   tar -xvf ${NV_WORKSPACE}/drive-linux_src/NVIDIA-kernel-module-source-TempVersion.tar.xz
   
   # Compile the display kernel module.
   export IGNORE_PREEMPT_RT_PRESENCE=1
   cd NVIDIA-kernel-module-source-TempVersion
   
   make modules \
      SYSSRC=$TEMP_DIR/kernel/$KERNEL_NAME \
      SYSOUT=$TEMP_DIR/kernel/out-linux \
      SYSSRCHOST1X=$TEMP_DIR/kernel/nvidia-oot/drivers/gpu/host1x/include \
      LD=${CROSS_COMPILE}ld.bfd \
      CC=${CROSS_COMPILE}gcc \
      AR=${CROSS_COMPILE}ar \
      CXX=${CROSS_COMPILE}g++ \
      OBJCOPY=${CROSS_COMPILE}objcopy \
      TARGET_ARCH=aarch64 \
      ARCH=arm64
   
   # Copy the display modules to the module_install path.
   make ARCH=arm64 -C $TEMP_DIR/kernel/out-linux M=$TEMP_DIR/kernel/NVIDIA-kernel-module-source-TempVersion/kernel-open INSTALL_MOD_STRIP=1 INSTALL_MOD_DIR=updates/opensrc-disp INSTALL_MOD_PATH=$TEMP_DIR/kernel/out-linux modules_install
   

Note

The steps (1–8) can also be run using:

NV_BUILD_KERNEL_CONFIG_NAME="defconfig" \
KERNEL_NAME=nvidia-6.1 \
$NV_WORKSPACE/drive-linux/kernel/compile_kernel.sh

9. Copy built Images and kernel modules to SDK default paths.

   # Copy the uncompressed (Image) kernel images to SDK kernel directory paths with the following command:
   export PROD_SUFFIX=<str> # "" for standard, "_prod" for production kernel
   rm -fv ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/images/*
   mkdir -p ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/images/
   cp -v $TEMP_DIR/kernel/out-linux/arch/arm64/boot/Image $TEMP_DIR/kernel/out-linux/vmlinux $TEMP_DIR/kernel/out-linux/System.map ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/images/
   
   # CAUTION: Before copying the new kernel images, back up the default kernels provided.
   rm -rf ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/modules/*
   mkdir -p ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/modules/
   cp -a $TEMP_DIR/kernel/out-linux/lib/modules/* ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/modules/
   

10. Copy built modules to root file system.

    Create a Build-FS JSON file named update_rfs.CONFIG.json at ${NV_WORKSPACE}/drive-linux/filesystem/.

    The FilesystemType parameter <fstype> should be set depending on the list of modules to be copied to the filesystem:

    • standard should be used for dev_nsr PCT.

    • production_debug should be used for test_nsr PCT.

    • production should be used for prod_nsr PCT.

    Example JSON:

    {
      "OS": "linux",
      "Output": "driveos-updated-rfs",
      "Base": "${NV_WORKSPACE}/drive-linux/filesystem/targetfs-images/${FILESYSTEM}_${DISTRO}_thor_rfs.img",
      "FilesystemType": "<fstype>",
      "CopyTargets": [
        "${COPYTARGETYAML_DIR}/copytarget-kernel-modules-6.1.thor.yaml"
      ],
      "PostInstalls": {
        "/etc/nvidia/run-once/nv-run-once-run-ldconfig": "target",
        "/etc/nvidia/run-once/nv-run-once-run-depmod": "target"
      }
    }
    

11. Rebuild the filesystem.

    export FILESYSTEM=dev_nsr_desktop # Assuming PCT is dev_nsr here. If PCT is test_nsr or prod_nsr, please set FILESYSTEM to test_nsr or prod_nsr respectively
    export DISTRO=ubuntu-24.04
    export NVRTKERNELNAME=$(basename ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/modules/*rt*-tegra)
    export NV_SDK_NAME_LINUX=drive-linux
    
    /usr/bin/python3 -B ${NV_WORKSPACE}/drive-common/filesystems/build-fs/18/bin/build_fs.py -w ${NV_WORKSPACE}/ -i ${NV_WORKSPACE}/drive-linux/filesystem/update_rfs.CONFIG.json -o ${NV_WORKSPACE}/drive-linux/filesystem/targetfs-images/
    

12. Update initramfs with built modules.

    # If Building a production kernel for prod_nsr/test_nsr PCTs, then please set:
    export INITRAMFS_IMAGE=${NV_WORKSPACE}/drive-linux/filesystem/initramfs/initramfs-nvidia-6.1.cpio
    
    # Else If building a standard kernel for dev_nsr PCT, please set:
    export INITRAMFS_IMAGE=${NV_WORKSPACE}/drive-linux/filesystem/initramfs/prod-initramfs-nvidia-6.1.cpio
    
    # Extract the initramfs.
    rm -rf ./initramfs; mkdir -p ./initramfs; cd ./initramfs
    cpio -imdu --quiet < ${INITRAMFS_IMAGE}
    
    # Run copytarget to copy kernel modules.
    export NVRTKERNELNAME=$(basename ${NV_WORKSPACE}/drive-linux/kernel/nvidia-6.1/preempt_rt${PROD_SUFFIX}/modules/*rt*-tegra)
    export NV_SDK_NAME_LINUX=drive-linux
    rm -rf lib/modules/*
    ${NV_WORKSPACE}/drive-common/filesystems/copytarget/1.4/copytarget.py $PWD ${NV_WORKSPACE}/ ${NV_WORKSPACE}/drive-linux/filesystem/copytarget/manifest/copytarget-initramfs.yaml --filesystem-type boot_initramfs
    
    # Create the depmod and alias
    for k in $(ls -d ./lib/modules/*); do
      echo "Running depmod for kernel modules at $k"
      depmod -b . -a $(basename $k)
    done
    

13. Rebuild updated initramfs.

    find . | cpio --quiet -H newc -o > ${INITRAMFS_IMAGE}
    

14. Pass the absolute path of the updated rootfs image using the bind_partitions command-line argument: LINUX_RFS_IMG_PATH=<absolute path of rootfs image> as seen in the example below.

    cd /drive/drive-foundation
    ./make/bind_partitions -b p3960-10-sw01 drive_av.linux -p dev_nsr LINUX_RFS_IMG_PATH=/drive/drive-linux/filesystem/targetfs-images/driveos-updated-rfs.img
    

15. Flash the target to the board using the SDK bootburn tool.

16. Boot the board. The built kernel modules will auto-load on boot.

• NVIDIA Supported Cross Toolchains
  • Rebuilding the Toolchain
• DriveOS Linux Yocto Components
  • Building the Yocto Project Components for NVIDIA DriveOS Thor
  • Flashing Yocto
  • Yocto Customizations
  • DM-Verity and Read-Only File System Support