> In article <20140128065536.gg1...@apb-laptoy.apb.alt.za>, > Alan Barrett <a...@cequrux.com> wrote: > >On Mon, 27 Jan 2014, Christos Zoulas wrote: > >>Modified Files: > >> src/sys/arch/i386/include: vmparam.h > >> > >>Log Message: > >>Cut down MAXDSIZE from 3G to 2.5G otherwise bottomup allocation ends up > >>supplying an out of bounds hint for sigcode (c001e000 > bf000000). Makes > >>a.out binaries work again. > > > >Will this make malloc fail 0.5GB earlier than before? The data > >size limits on i386 are already annoyingly small, and I would > >prefer not to make them smaller. Please could you find a way to > >penalise only a.out programs. > > I don't think it could allocate 3G before either. I think that the data > segment would smash into the stack then. I can test though to verify.
it probably doesn't matter since most of the memory you'll end up using for a large process with be counted against RLIMIT_AS not RLIMIT_DATA, with modern malloc(3). .mrg.