MAP_POPULATE vs. MADV_HUGEPAGES

From	Avi Kivity <avi@scylladb.com>
Newsgroups	linux.kernel
Subject	MAP_POPULATE vs. MADV_HUGEPAGES
Date	2017-03-15 18:00 +0100
Message-ID	<tlkrg-1Wi-19@gated-at.bofh.it> (permalink)
Organization	ScyllaDB

Show all headers | View raw

A user is trying to allocate 1TB of anonymous memory in parallel on 48 
cores (4 NUMA nodes).  The kernel ends up spinning in 
isolate_freepages_block().


I thought to help it along by using MAP_POPULATE, but then my 
MADV_HUGEPAGE won't be seen until after mmap() completes, with pages 
already populated.  Are MAP_POPULATE and MADV_HUGEPAGE mutually exclusive?


Is my only option to serialize those memory allocations, and fault in 
those pages manually?  Or perhaps use mlock()?

Back to linux.kernel | Previous | Next — Next in thread | Find similar | Unroll thread

Thread

MAP_POPULATE vs. MADV_HUGEPAGES Avi Kivity <avi@scylladb.com> - 2017-03-15 18:00 +0100
  Re: MAP_POPULATE vs. MADV_HUGEPAGES Michal Hocko <mhocko@kernel.org> - 2017-03-16 13:40 +0100
    Re: MAP_POPULATE vs. MADV_HUGEPAGES Avi Kivity <avi@scylladb.com> - 2017-03-16 14:30 +0100
      Re: MAP_POPULATE vs. MADV_HUGEPAGES Michal Hocko <mhocko@kernel.org> - 2017-03-16 15:50 +0100
        Re: MAP_POPULATE vs. MADV_HUGEPAGES Avi Kivity <avi@scylladb.com> - 2017-03-16 16:00 +0100
          Re: MAP_POPULATE vs. MADV_HUGEPAGES Michal Hocko <mhocko@kernel.org> - 2017-03-16 16:10 +0100

csiph-web