[Michael Biebl] > Second, how should this above case be handled? I guess it's typical > for fuse based filesystems, that there is a running process as long as > the corresponding partition is mounted. What is the correct fix: > 1.) Should sendsigs omit those (mount) processes by default?
This is a bad idea, as sendsigs would have to accumulate special handling when there already is a generic mechianism to avoid being killed, the omitpid feature. > 2.) Should those fuse filesystems utilize and set an entry in > /lib/init/rw/sendsigs.omit.d/? > 3.) Should fuse filesystems be unmounted before sendsigs? Either of these make sense to me. Only those processes needed when umounting /usr/ should survive sendsigs. But is this fuse file system umounted by umountnfs? If so, it might make sense to update the omitpid list with it. > 4.) Should the (fuse) mount programs react on SIGTERM? Could be. Leave that to someone who know the system. Happy hacking, -- Petter Reinholdtsen _______________________________________________ Pkg-sysvinit-devel mailing list [email protected] http://lists.alioth.debian.org/mailman/listinfo/pkg-sysvinit-devel
