https://bugzilla.novell.com/show_bug.cgi?id=466484 User werner@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=466484#c104 Dr. Werner Fink <werner@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |mboman@novell.com --- Comment #104 from Dr. Werner Fink <werner@novell.com> 2009-03-11 05:00:32 MST --- mkill only opens /proc/mounts to get all active mount points and /proc to read the directory, then it uses readlink(2), open(2) and opendir(3) to determine which running program makes a mount point busy. And running strace -e open,readlink,close mkill -0 /dev does not show a file descriptor leak. Beside this mkill does not stop /sbin/udevd nor programs which have /dev/fuse open. The last one because if those would be terminated the underlying fuse file system becomes dirty. Could it be that there is a program which opens it own fuse device which the program its self creates with makedev(3) and mknod(2)? Or could it be that there is a program which uses a /dev/fuse within a chroot environment ... maybe a combination with a network based file system (samba,NFS) and a local file system. Nevertheless the new mkill sorts kill(pid,SIGTERM) in the reverse order of the mount points found in /proc/mounts. Or implies the order of your mounts that some of the mount points remains busy? To see this you should compare /etc/mtab and /proc/mounts. I'll add Magnus which maybe can explain what happens with this gvfs-fuse-daemon which seems to hold /home/SSIS/diego.ercolani/.gvfs or /home busy or dirty. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.