> > 1st read after client's reboot - at client's harddisk speed via > > cachefs > > Rebooting is actually unnecessary here. All you need do is unmount the NFS > filesystem (and possibly the cache too). I quite agree, I just thought that verification with reboot is more immune to possible human errors a tester can make (as each of them is using it the first time) and might also verify other aspects of the kernel code itself (I guess it might be easy to leave some kernel structures hanging around even after those two unmounts and have a code which works after unmounts/mounts but does not work after reboot). Thank's for all your work, David! Regards Vaclav
