On Wed, Apr 23, 2025 at 03:15:47AM +0100, Al Viro wrote:
> On Tue, Apr 22, 2025 at 03:53:43PM -0400, Eric Chanudet wrote:
>
> > I'm not quite following. With umount -l, I thought there is no guaranty
> > that the file-system is shutdown. Doesn't "shutdown -r now" already
> > risks loses without any of these changes today?
>
> Busy filesystems might stay around after umount -l, for as long as they
> are busy. I.e. if there's a process with cwd on one of the affected
> filesystems, it will remain active until that process chdirs away or
> gets killed, etc. Assuming that your userland kills all processes before
> rebooting the kernel, everything ought to be shut down, TYVM...
>
> If not for that, the damn thing would be impossible to use safely...
>
Right, that ties up with Christian's earlier reply and was also stated
in 9ea459e110df ("delayed mntput") description.
Thanks for your patience and explanations.
--
Eric Chanudet
