Link NVIDIA driver libraries into a path the transcoder loader searches - #172
Open
magnavoid wants to merge 1 commit into
Open
Link NVIDIA driver libraries into a path the transcoder loader searches#172magnavoid wants to merge 1 commit into
magnavoid wants to merge 1 commit into
Conversation
The Plex Transcoder is loaded by Plex's bundled musl dynamic loader,
which does not read /etc/ld.so.cache and only probes a fixed set of
directories. The NVIDIA container toolkit mounts the host's driver
libraries at whatever path the host distro uses; on Debian hosts that is
/usr/lib/<multiarch>/nvidia/current, which is only discoverable through
the glibc linker cache. The transcoder's dlopen("libcuda.so.1") then
fails and hardware transcoding silently falls back to software, logging
only a misleading 'opening hw device failed ... Operation not permitted'
even though nvidia-smi works inside the container.
Resolve the injected libraries from the linker cache (the toolkit's own
record of what it mounted, populated by the ldconfig it runs at
container setup) and symlink them into /usr/local/lib, which the bundled
loader searches. No-ops when the NVIDIA runtime is not in use, never
overwrites real files, and refreshes stale links after host driver
upgrades.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The Plex Transcoder is loaded by Plex's bundled musl dynamic loader,
which does not read /etc/ld.so.cache and only probes a fixed set of
directories. The NVIDIA container toolkit mounts the host's driver
libraries at whatever path the host distro uses; on Debian hosts that is
/usr/lib//nvidia/current, which is only discoverable through
the glibc linker cache. The transcoder's dlopen("libcuda.so.1") then
fails and hardware transcoding silently falls back to software, logging
only a misleading 'opening hw device failed ... Operation not permitted'
even though nvidia-smi works inside the container.
Resolve the injected libraries from the linker cache (the toolkit's own
record of what it mounted, populated by the ldconfig it runs at
container setup) and symlink them into /usr/local/lib, which the bundled
loader searches. No-ops when the NVIDIA runtime is not in use, never
overwrites real files, and refreshes stale links after host driver
upgrades.