On Fri, 2005-03-25 at 11:54 +0800, Ow Mun Heng wrote: > On Fri, 2005-03-25 at 10:28 +0800, Ow Mun Heng wrote: > > On Thu, 2005-03-24 at 14:11 -0600, Aleksandar Milivojevic wrote: > > > William Hooper wrote: > > > > Ow Mun Heng said: > > > > [snip] > > > > > > > >>The author then compares the results with that of BSD eg: OpenBSD and > > > >>FreeBSD IIRC, and notes that on a system with Pentium 90 w/ 32MB Ram, on > > > >>using the forkbomb, doesn't bring the system down to it's knees. > > > > > > > > > > > > Following the discussion on the -devel list, the BSDs aren't immune. > > > > > > > > https://www.redhat.com/archives/fedora-devel-list/2005-March/msg01201.html > > > > > > I gave a try to "while(1) fork();" on OpenBSD (SS5, 64MB RAM). You > > > can't go much simpler than that. Results: > > > > > > Running as normal user, max number of processes is limited to 64 by > > > default, load average went to 50-something. Perfectly able to kill > > > offending process(es). > > > > > > Running as root, max number of processes is limited to 340, load average > > > went to almost 300. Not able to do much, since root couldn't create any > > > new processes (in order to kill existing), but top worked, and already > > > running processes worked. Anyhow, if somebody does fork bomb as root, > > > he could just as well type "halt" or wipe out the disk or do some other > > > irreversable damage. > > > > > > The maximum number of processes per user is hard coded in OpenBSD > > > kernel. It is 340 for root, and 64/128 for normal users in my generic > > > kernel (soft limit is 64, hard limit is 128). If you want more either > > > for root or normal users, you need to recompile the kernel. In *BSD > > > world, recompiling the kernel is something you do twice a day anyhow, so > > > this isn't much of a problem ;-) > > > > > > For those interested, these are ulimits on stock OpenBSD (user and root, > > > soft and hard). > > > > > [SNIP] > > > > I see.. So, ulimits _are_ set by default(lower) in *BSDs and not as what > > I thought that they had some kind of process watchdog in place. > > > > One more thing, what about if there are 20 users and all the 20 users > try to fork at the same time? What will happen then? > Obviously each will fork to his maximum limit unless the system process table fills up or the whole system grinds to a halt due to load/resource contention.
