On Wed, Jun 14, 2006 at 01:58:46PM -0400, John Baldwin wrote: > On Wednesday 14 June 2006 13:53, Paul Allen wrote: > > From Jason Evans <jasone_at_freebsd.org>, Wed, Jun 14, 2006 at > 10:35:03AM -0700: > > > This is on a 32-bit system, right? If so, what's happening is that the > > > brk-managed space (data segment) is being fragmented, such that the > > > address space isn't returned to the OS. However, this is not really a > > > memory leak, since madvise() is called in order to let the kernel know > > > that the unused space need not be swapped out. > > > > And in particular this should manifest itself as 'RES' declining but 'SIZE' > > in top remaining unchanged. Until an over-commit hating individual manages > > to get a patch into the tree, this should not be a concern at all. > > And any such patch would have to have a knob that defaulted to keeping > overcommit on anyway. :-) Ha ! Exactly. My patch has such knob :). Shameless plug: http://kostikbel.narod.ru/overcommit Testing the patch on recent CURRENT, I did notice that jemalloc really likes the swap :).
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:57 UTC