* Update nvidia-container-toolkit, libnvidia-container to v1.16.2
* Refresh nvidia-container-toolkit alt-roots patch
* Drop obsolete backported patch and config.toml template file
* Use nvidia-modprobe patch in the libnvidia-container sources
instead of maintaining our own copy
* Other minor cleanup of obsolete settings in libnvidia-container
recipe
* Fix PACKAGE_ARCH of libnvidia-container for tegra platforms
* Add missing BSD-3-Clause license to nvidia-container-toolkit recipe
* Update license checksums in nvidia-container-toolkit recipe;
no license-affecting changes.
Signed-off-by: Matt Madison <matt@madison.systems>
When device-tree builds moved to nvidia-kernel-oot, the dtb files built from
the platform dts files were getting stashed into the sysroot area for future
use during image build, but the dts files themselves were *not* saved anywhere.
This broke external device-tree recipes that wanted to include the platform dts
and add modifications.
Add export of dts files into nvidia-kernel-oot sysroot area,
under usr/src/device-tree subdir.
Signed-off-by: S. Lockwood-Childs <sjl@vctlabs.com>
- 'mv generate_bup_payload.sh doflash.sh' should be
conditional on TEGRA_SIGNING_EXCLUDE_TOOLS. Those who
use external signing servers will set this to "1" which
skips the generation of generate_bup_payload.sh. See
invocations of tegraflash_generate_bupgen_script in
image_types_tegra.bbclass.
Signed-off-by: Chad McQuillen <chad.mcquillen@lexmark.com>
nvcc needs the cicc binary provided in the cuda-nvvm package, so
make sure that gets installed with the host-side SDK tools.
Signed-off-by: Matt Madison <matt@madison.systems>
zstd is a lot faster when compressing and decompressing, which helps
with iteration speed.
Use the CONVERSIONTYPES machinery builtin rather than re-implementing it
ourselves here.
Signed-off-by: Austin Schuh <austin.linux@gmail.com>
Signed-off-by: Matt Madison <matt@madison.systems>
OE-Core master has udpated how the toolchain virtual providers
are handled, so update the layer accordingly.
This affects the gcc-for-nvcc extra toolchain build we have for
CUDA builds, and a few other recipes that have some explicit
references to the old toolchain virtual/<something> names.
Signed-off-by: Matt Madison <matt@madison.systems>
USB instances might not be consistent between the Jetson boot ROM
and the mass storage USB gadget in the running OS, so make the
cross-check of the instance ID optional, and off by default.
This means that when flashing secureboot-enabled devices using
pre-signed binaries, only one target can be flashed at a time
from a given host.
Signed-off-by: Matt Madison <matt@madison.systems>
* Drop OpenSSL 3 support patch now included upstream
* Rework OE build integration to be a bit cleaner
* Update to use BBCLASSEXTEND for native, instead of
having a direct -native recipe
Signed-off-by: Matt Madison <matt@madison.systems>
Fix the change introduced in ff0795338b
to use a relative path, rather than an absolute path, so the full
build path doesn't get embedded in any of OP-TEE packages.
Signed-off-by: Matt Madison <matt@madison.systems>
Since we can't get the target's serial number out of the CVM EEPROM when
using pre-signed binaries, look for the USB instance in the device path
as a way of cross-checking that a storage device matches the target we're
trying to flash.
Signed-off-by: Matt Madison <matt@madison.systems>
* Tweak the flashing script and XML file naming for the pre-signed case
as is done when the script runs the signing itself. Note that external
flash layouts are assumed not to require any tegrasign-type signing, which
should always be the case for R36.x and later.
* Make the script tolerant of not being able to fetch the target's serial
number from the CVM EEPROM. There doesn't appear to be a way to fetch
the CVM contents with pre-signed boot firmware.
With these changes, when using pre-signed binaries, only one target can
be flashed at a time from a given host.
Signed-off-by: Matt Madison <matt@madison.systems>
The odmsign functions do the same post-processing for RCM boot signing
as they do for main signing, so be sure to save and restore the main
files after the second call for signing the RCM boot blob.
Signed-off-by: Matt Madison <matt@madison.systems>
Now that we have patched the NVIDIA out-of-tree drivers to
be compatible with this version, we can update.
Signed-off-by: Matt Madison <matt@madison.systems>
With the repos hosted or mirrored on Github so we can more
easily manage patch additions for kernel versions beyond what
NVIDIA provides in a given Jetson Linux release.
Signed-off-by: Matt Madison <matt@madison.systems>
With L4T R36.4, the odmsign.func script now supports signing
the RCM boot blob used for the initrd-based flashing process.
While the initrd-flash script was updated to handle this, more
changes are needed to handle signing all binaries during the
build.
* Update the flash helper script to run signing on the RCM
boot blob when needed.
* Update the initrd-flash script to drop the extra helper
invocation for signing the RCM boot blob, since that is
now covered with the normal signing step for the internal
flash. And when pre-signed, initrd-flash just looks for
the RCM boot blob in the subdirectory now created during
build-time signing.
Signed-off-by: Matt Madison <matt@madison.systems>
We want to be able to pull in the module load configuration by default
but need to be careful to not enable the 'modeset' parameter on the
driver unless we won't be using x11 in the image. We can't check
distro features, so we make the modprobe conf RCONFLICT with x11, and
don't allow it to be pulled in if 'modeset' is enabled.
Also add variables to allow overriding the setting for the 'modeset'
and 'fbdev' parameters in a bbappend, local.conf file, or similar.
Signed-off-by: Kurt Kiefer <kekiefer@gmail.com>
The nvidia-drm kernel module and associated load config are needed
by libdrm but were only getting picked up by weston-init.
There are valid configurations (headless, egl-only) that need
libdrm but do not pull in weston, so the runtime recommendation
is moved to the libdrm bbappend.
Signed-off-by: Kurt Kiefer <kekiefer@gmail.com>
sinpi/cospi/sinpif/cospif are now defined in glibc as part of C23 extentions
but they conflict with CUDA declrations, therefore use the glibc definitions
when its provided by glibc
Fixes https://github.com/OE4T/meta-tegra/issues/1784
Signed-off-by: Khem Raj <raj.khem@gmail.com>
When flashing a board fused with PKC hash but no SBK, we should not
supply a SBK key as that would encrypt the flash contents which would
then not be able to be read by the board when booting, as the board
won't know the SBK key.
Signed-off-by: Andrew Bradford <andrew.bradford@konsulko.com>
With the R36 series, the nvidia-l4t-gstreamer package now comes off
the 'common' feed instead of the per-SoC feed, which affects multiple
recipes.
Add a new variable, L4T_DEB_IS_COMMON, to the include file we use for
the BSP recipes, to let them specify that the main package should come
from the common feed. Update the recipes for the binary-only gstreamer
plugins, nvdsseimeta, and the DLA compiler to use this new mechanism.
Signed-off-by: Matt Madison <matt@madison.systems>
