This is largely a placeholder for now. The tl;dr is the upstream lvmteam folks (who control the reference impllementation of the cookie sync protocol in libdevmapper) are working on a modernisation that will eliminate the current SysV semaphore based sync implementation in favour of native UDEV monitor threads (don't ask me for more details rn — I don't have them).
My current plan is to extend dm_udev_sync.rs with the New Way™ but keep the existing implementation to allow a reasonable transition period. We can then use either runtime detection (my preference), or a feature flag to select which to use in any given situation. I believe this will allow us to navigate the migration with minimal fallout.
I will update this issue as more details emerge.
This is largely a placeholder for now. The tl;dr is the upstream lvmteam folks (who control the reference impllementation of the cookie sync protocol in
libdevmapper) are working on a modernisation that will eliminate the current SysV semaphore based sync implementation in favour of native UDEV monitor threads (don't ask me for more details rn — I don't have them).My current plan is to extend
dm_udev_sync.rswith the New Way™ but keep the existing implementation to allow a reasonable transition period. We can then use either runtime detection (my preference), or a feature flag to select which to use in any given situation. I believe this will allow us to navigate the migration with minimal fallout.I will update this issue as more details emerge.