Messages in this thread | | | Date | Thu, 17 Jun 1999 11:56:56 +0200 | From | Gabor Lenart <> | Subject | Re: [patch] 2.2.10_andrea-VM6.gz |
| |
On Thu, Jun 17, 1999 at 03:36:50AM +0200, Andrea Arcangeli wrote: > I also changed how I choose the task to kill: when 2.2.10_andrea-VM6 > reaches OOM, it sends a SIGKILL to all tasks that belongs to the bigger MM > in the system. I perfectly know it's far from being a perfect solution
And what about X server or something like this ? Killing it has one disadvantage : the console stuck in undefined state. So it might be good to send something other signal than SIGKILL, and after SIGKILL some seconds or like ... (or detect that a process ignore the first sent signal (SIG_IGN?), and if so, send SIGKILL for it without any delay). OK, maybe my knowledge on kernel internals is very limited to find solution for this problem but killing an X server with SIGKILL is not a very good solution imho.
Some months ago somebody at our university was talking about implement a user space process 'guard' daemon with mlock()ed memory which cheks system state, resource usage per user basis etc etc and sysadm defined actions declared in a config file for example. Maybe we only need a little kernel interface and a user mode program to do this task in the right way ?
-- ---[ LGB/DC ]------------[ University Of Veszprém ]------[ Lénárt Gábor ]--- Ready to love, ready to fly. Ready to fail, ready to flow, ready to die ... ---[ 88/422022(4602) ]---[ http://lgb.hal.vein.hu ]---------[ 87/477074 ]--- finger lgb@hal2000.hal.vein.hu for more information including my PGP&GEEK code
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |