@otto This sounds like a really interesting idea. Definitely sounds like the next step in terms of OpenBSD memory handling.
@otto So part of this idea is in the new MAP_CONCEAL. None of the wiping stuff but it gets kept out of core dumps.
@kurtm wiping is done when a page is unmapped. That's already the case. The userland malloc part will take care of setting the conceal flag on pages allocated by the special function. Small allocations (that share a page) will be wiped on free, large allocations will be unmapped on free. Reallocate of a special allocation will retain its special status.
@otto so looking forward to take a look at that.