Per the suggestion from Michal Hocko [1], DEFERRED_STRUCT_PAGE_INIT requires some ordering wrt other initialization operations, e.g. page_ext_init has to happen after the whole memmap is initialized properly. For SPARSEMEM this requires to wait for page_alloc_init_late. Other memory models (e.g. flatmem) might have different initialization layouts (page_ext_init_flatmem). Currently DEFERRED_STRUCT_PAGE_INIT depends on MEMORY_HOTPLUG which in turn depends on SPARSEMEM || X86_64_ACPI_NUMA depends on ARCH_ENABLE_MEMORY_HOTPLUG
and X86_64_ACPI_NUMA depends on NUMA which in turn disable FLATMEM memory model: config ARCH_FLATMEM_ENABLE def_bool y depends on X86_32 && !NUMA
so FLATMEM is ruled out via dependency maze. Be explicit and disable FLATMEM for DEFERRED_STRUCT_PAGE_INIT so that we do not reintroduce subtle initialization bugs
[1] http://lkml.kernel.org/r/20160523073157.GD2278@dhcp22.suse.cz
CC: Michal Hocko mhocko@kernel.org Signed-off-by: Yang Shi yang.shi@linaro.org --- v2: Adopted Michal's comments for the commit log
mm/Kconfig | 1 + 1 file changed, 1 insertion(+)
diff --git a/mm/Kconfig b/mm/Kconfig index 2664c11..22fa818 100644 --- a/mm/Kconfig +++ b/mm/Kconfig @@ -649,6 +649,7 @@ config DEFERRED_STRUCT_PAGE_INIT default n depends on ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT depends on MEMORY_HOTPLUG + depends on !FLATMEM help Ordinarily all struct pages are initialised during early boot in a single thread. On very large machines this can take a considerable