mm: disable NUMA_BALANCING_DEFAULT_ENABLED and TRANSPARENT_HUGEPAGE on PREEMPT_RT
authorSebastian Andrzej Siewior <bigeasy@linutronix.de>
Fri, 5 Nov 2021 20:35:27 +0000 (13:35 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 6 Nov 2021 20:30:33 +0000 (13:30 -0700)
TRANSPARENT_HUGEPAGE:
  There are potential non-deterministic delays to an RT thread if a
  critical memory region is not THP-aligned and a non-RT buffer is
  located in the same hugepage-aligned region. It's also possible for an
  unrelated thread to migrate pages belonging to an RT task incurring
  unexpected page faults due to memory defragmentation even if
  khugepaged is disabled.

Regular HUGEPAGEs are not affected by this can be used.

NUMA_BALANCING:
  There is a non-deterministic delay to mark PTEs PROT_NONE to gather
  NUMA fault samples, increased page faults of regions even if mlocked
  and non-deterministic delays when migrating pages.

[Mel Gorman worded 99% of the commit description].

Link: https://lore.kernel.org/all/20200304091159.GN3818@techsingularity.net/
Link: https://lore.kernel.org/all/20211026165100.ahz5bkx44lrrw5pt@linutronix.de/
Link: https://lkml.kernel.org/r/20211028143327.hfbxjze7palrpfgp@linutronix.de
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
init/Kconfig
mm/Kconfig

index 11f8a84..21b1f48 100644 (file)
@@ -901,7 +901,7 @@ config NUMA_BALANCING
        bool "Memory placement aware NUMA scheduler"
        depends on ARCH_SUPPORTS_NUMA_BALANCING
        depends on !ARCH_WANT_NUMA_VARIABLE_LOCALITY
-       depends on SMP && NUMA && MIGRATION
+       depends on SMP && NUMA && MIGRATION && !PREEMPT_RT
        help
          This option adds support for automatic NUMA aware memory/task placement.
          The mechanism is quite primitive and is based on migrating memory when
index d16ba92..9f1e009 100644 (file)
@@ -371,7 +371,7 @@ config NOMMU_INITIAL_TRIM_EXCESS
 
 config TRANSPARENT_HUGEPAGE
        bool "Transparent Hugepage Support"
-       depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE
+       depends on HAVE_ARCH_TRANSPARENT_HUGEPAGE && !PREEMPT_RT
        select COMPACTION
        select XARRAY_MULTI
        help