From: Moreno B. <ba...@de...> - 2005-01-31 10:03:21
|
On Fri, 2005-01-28 at 20:54, Brian Haney wrote: > Hmmm. Thanks for explaining that to me. I see in $LINUXSRC/hpc/hpcproc.c > that proc_mosix_pid_set_lock() checks for (task == current) before calling > mig_set_lock(), (which, looking at mig.c, will work only for task == > current), but if I just commented out the following in > proc_mosix_pid_set_lock(): > > if(lock) > return(-EINVAL); > > it seems that lines immediately following would allow superuser to lock an > arbitrary process. It's not enough, those lines allow superuser to UN-lock an arbitrary process. In order to enable locking you should use something like this http://www.democritos.it/~baro/sw/lock/lock.patch It works only on locally running processes (never migrated or migrated back to home node), "current" dependencies complicate the locking of migrated ones (but I may wrong). I'm not sure it's safe. Although it works for me, try it at your owk risk. Ciao Moreno |