On Thu, Jan 08, 2009 at 02:59:45PM -0600, Jonathan Brassow wrote: > Ummm? Here is the patch: https://www.redhat.com/archives/dm-devel/2009-January/msg00058.html > > brassow > > On Jan 8, 2009, at 1:25 PM, Konrad Rzeszutek wrote: > >> >> On memory constrained systems, where a large amount of LUNs are >> assigned (say more enough to make multipath blossom to 70% of available >> memory space) forking would fail returning -ENOMEM. The kernel has enough >> free pages for the new process, but when it clones the new child process >> from the parent process space it accounts the parent process heap space >> against the child (which counts for more pages than there are free pages). >> >> Using vfork passes a flag to clone which will inhibit this calculation >> and instead check for free pages. The danger in using vfork is that the >> child >> could use the values from the parent, but fortunatly we don't do that and >> immediately call 'execve' swipping the process space free. >> >> >> -- >> dm-devel mailing list >> dm-devel@xxxxxxxxxx >> https://www.redhat.com/mailman/listinfo/dm-devel > > -- > dm-devel mailing list > dm-devel@xxxxxxxxxx > https://www.redhat.com/mailman/listinfo/dm-devel -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
