@icewind1991 looks like "$view->removeMount()" doesn't lock.
We should exclusive-lock the mount point path (not the storage root) to avoid concurrent processes to access that path during the umount (in case sharing takes a while...)
Requires #17070 to be merged first for the extra argument.
The same should be done for adding mount points (when sharing, etc)
@cmonteroluque @karlitschek @DeepDiver1975 moving to 8.1.1.
Needed to make locking more watertight.
@PVince81 Is there any PR for this?
@PVince81 what is the criticality of this?
@PVince81 agree that it needs to be better, 8.1.2 not looking possible at this point.
Not sure why we put this on 8.1.4. Other remaining locking tasks like #17243 are now set to 9.0.
I'm not sure this is needed, I don't think there is any chance for data loss since the worst that can happen is a different process writing to a mount that is being removed, which is the same result as when the other processed finished it's write before the storage is removed
Best would be to do a test with sleep() in the right places to make sure it works as you think it does.
@nickvergessen mind taking over ? If we find that we don't need a lock there then I'm fine dropping it, but we need to verify this.