Commit Graph

299 Commits

Author SHA1 Message Date
Evan Lezar
bdf997c761 Use semver package to parse CUDA version
This avoids the use of scanf on a user-provided string which is flagged
as a security vulnerability.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-22 14:57:52 +02:00
Evan Lezar
c224832a6d Add log-level config option for nvidia-container-runtime
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 13:56:17 +02:00
Evan Lezar
5211960fc3 Merge branch 'detect-gpus-flag' into 'master'
Detect use of --gpus flag in experimental mode

See merge request nvidia/container-toolkit/container-toolkit!125
2022-04-08 11:18:11 +00:00
Evan Lezar
45160b88a4 Remove exsiting NVIDIA Container Runtime Hooks from the spec
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 12:03:22 +02:00
Evan Lezar
a9a4704273 Raise error if hook invoked in experimental mode without force flag
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 12:03:22 +02:00
Evan Lezar
7ec3cd0b5b Fix creation of CSV parser in create-symlinks
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 11:39:18 +02:00
Evan Lezar
ab7f25500f Fix creation of CSV parser in create-symlinks
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 11:36:48 +02:00
Evan Lezar
196d5c5461 Move NVIDIA Container Runtime Hook executable name to shared constant
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 11:29:27 +02:00
Evan Lezar
f07d110e85 Use DefaultExecutableDir to determine default paths
This change adds a DefaultExecutableDir = /usr/bin constant that is used
to construct default paths for executables instead of specifying these
explicitly.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 11:28:03 +02:00
Evan Lezar
1ebd48dea6 Merge branch 'add-symlink-hook' into 'master'
Add hook create-symlinks subcommand to create symlinks in container

See merge request nvidia/container-toolkit/container-toolkit!121
2022-04-08 09:14:07 +00:00
Evan Lezar
f7c74d35cc Merge branch 'add-hooks-cli' into 'master'
Add nvidia-ctk CLI with hook command and update-ldcache subcommand to update LD cache

See merge request nvidia/container-toolkit/container-toolkit!115
2022-04-08 09:13:39 +00:00
Evan Lezar
67602b28f9 Return unmodified runtime if specModifier is nil
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-08 07:50:40 +02:00
Evan Lezar
907736b053 Inject symlinks hook for creating symlinks in a container
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 20:25:55 +02:00
Evan Lezar
ecb4ef495a Add create-symlinks subcommand to create symlinks in container for specified CSV files
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 20:25:55 +02:00
Evan Lezar
95797a8252 Move reading of container state for internal/oci package
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 20:25:55 +02:00
Evan Lezar
c87ae586d4 FIX: Rename containerSpec flag to container-spec
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 20:25:19 +02:00
Evan Lezar
d970d0a627 Add discovery for ldconfig hook that updates the LDCache
This change adds a discovered hook for updating the ldcache as a container-create
hook. The mounts from a discoverer are inspected to determine the folders that must
be added to the cache using the nvidia-ctk hook update-ldcache command.

This is added to the "csv" discovery mode for the experimental runtime.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 20:25:19 +02:00
Evan Lezar
1c892af215 Add hook command to nvidia-ctk with update-ldcache subcommand
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 16:38:01 +02:00
Evan Lezar
c945cc714d Add stub nvidia-ctk CLI
This change adds an nvidia-ctk CLI that is used as the basis for
utilities related to the NVIDIA Container Toolkit.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 16:32:25 +02:00
Evan Lezar
bb086d4b44 Add auto discover mode and use this as the default
This change adds an 'auto' discover mode that attempts to select the correct mode
for a given platform. This currently attempts to detect whether the platform is a
Tegra-based system in which case the 'csv' discover mode is used. The 'legacy'
discover mode is used as the fallback.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 15:37:03 +02:00
Evan Lezar
26d2873bb2 FIX: Rename DefaultRoot to DefaultMountSpecPath
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 14:11:52 +02:00
Evan Lezar
6682bc90b4 Add support for NVIDIA_REQUIRE_JETPACK envvar
This change ensures that by default, the CSV discovery only considers the base CSV
files (l4t.csv, drivers.csv, devices.csv) and skips the rest unless the
NVIDIA_REQUIRE_JETPACK is set to "csv-mounts=all", in which case, all CSV files in the
specified folder are considered.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 10:46:26 +02:00
Evan Lezar
1c05a463bd Add csv discovery mode to experimental runtime
This change adds support for a "csv" discovery mode to the experimental runtime.
If this is set with experimental = true, a CSV-based discovery of devices and
mounts are used to define the modifications required to the OCI spec. The edits
are expressed as CDI ContainerEdits.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-07 10:45:19 +02:00
Evan Lezar
9ce690093d FIX: Make isNVIDIAContainerRuntimeHook mode idiomatic
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-06 17:18:06 +02:00
Evan Lezar
b8dd473343 FIX: Simplify hook remover
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-06 17:15:57 +02:00
Evan Lezar
11aa1d2a7d FIX: Factor out specModifier construction into function
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-06 15:18:12 +02:00
Evan Lezar
e6730fd0f0 FIX: Don't log that hooks is being removed if it is not
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-06 15:13:32 +02:00
Evan Lezar
8db287af8b FIX: Fix typo in comment
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-06 14:46:27 +02:00
Evan Lezar
282a2c145e Fix typo in variable name
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:26 +02:00
Evan Lezar
d0608844dc Add basic README for nvidia-container-runtime
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:26 +02:00
Evan Lezar
a26d02890f Make error logging less verbose by default
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:26 +02:00
Evan Lezar
14fe35c3f4 Implement hook remover for existing nvidia-container-runtime-hooks
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:26 +02:00
Evan Lezar
d12dbd1bef Read top-level config to propagate Root to experimental runtime
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:25 +02:00
Evan Lezar
239b6d3739 Implement experimental modifier for NVIDIA Container Runtime
This change enables the experimental mode of the NVIDIA Container Runtime. If
enabled, the nvidia-container-runtime.discover-mode config option is
queried to determine how required OCI spec modifications should be defined.
If "legacy" is selected, the existing NVIDIA Container Runtime hooks is
discovered and injected into the OCI spec.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:25 +02:00
Evan Lezar
7137f4b05b Move runtime config to internal package
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:24 +02:00
Evan Lezar
0c7eb93d62 Add experimental option to NVIDIA Container Runtime config
This change adds an experimental option to the NVIDIA Container Runtime config. To
simplify the extension of this experimental mode in future an error is raised if
this is enabled.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-04-04 14:16:24 +02:00
Evan Lezar
c6dfc1027d Move modifier code for inserting nvidia-container-runtime-hook to separate package
This change moves the code defining the insertion of the nvidia-container-runtime
hook to a separate package. This allows for better distinction between the existing
and experimental modifications.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-03-29 11:05:42 +02:00
Evan Lezar
4177fddcc4 Import modifying runtime abstraction from experimental runtime
This change imports the modifying runtime abstraction from the
experimental branch. This encapsulates the checks for whether
modification is required, and forwards the loaded spec to
the specified modifier. This allows for the same code to be
reused when performing more complex modifications.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-03-29 11:05:42 +02:00
Evan Lezar
bf8c3bab72 Add test package with GetModuleRoot and PrependToPath function
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-03-29 11:05:41 +02:00
Evan Lezar
c5c2ffd68f Ensure that Exec error is also logged to file
This change removes unneeded logging and renames the return error value to rerr
to avoid it being aliased by local error values.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2022-03-29 11:05:41 +02:00
Evan Lezar
0a2db7c70e Add nvidia-container-config option to overide drivercapabilities
This change adds support for a supported-driver-capabilities config
option in the config.toml file that allows the driver capabilities
associated with the NVIDIA_DRIVER_CAPABILITIES=all environment variable.
This can be used on platforms such as Jetson to remove unsupported
capabilities such as "ngx".

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-11-17 16:53:08 +01:00
Evan Lezar
ec8a6d978d Import cmd/nvidia-container-runtime from experimental branch
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-10-29 16:49:05 +02:00
Evan Lezar
5996379fcc Add changelog entry for config.json path changes
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-10-26 14:29:28 +02:00
wenjun gao
350c8893fb skip error when bundleDir not exist 2021-10-26 17:51:12 +08:00
Evan Lezar
10cd42273e Update package references
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-09-07 13:13:03 +02:00
Evan Lezar
b6a585c77d Copy code from nvidia-container-runtime
This change copies the cmd/nvidia-container-runtime, internal, and test
folders from github.com/NVIDIA/nvidia-container-runtime@8a63b4b34f3ce3b4167f0516aa3f7207ca280dfb

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-09-07 13:13:03 +02:00
Evan Lezar
c2ac6db43b Revert "Add support for NVIDIA_FABRIC_DEVICES"
This reverts commit f828efcf64.
2021-08-18 15:17:59 +02:00
Evan Lezar
f828efcf64 Add support for NVIDIA_FABRIC_DEVICES
This change adds support for the NVIDIA_FABRIC_DEVICES envvar. The (non-empty)
value of this envvar is passed to the NVIDIA Container CLI using the --fabric-device
command line flag and allows for nvswitch and nvlink devices to be mounted
into the container.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-08-11 10:33:55 +02:00
Evan Lezar
f72b79cc2a Move pkg to cmd/nvidia-container-toolkit
This change moves the pkg folder to `cmd/nvidia-container-toolkit` to
better match go best practices. This allows, for example, for the
`cmd/nvidia-container-toolkit` to be go installed.

The only package included in `pkg` was `main`.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-06-08 15:20:59 +02:00