-config MEMORY_HOTPLUG_DEFAULT_ONLINE
- bool "Online the newly added memory blocks by default"
- depends on MEMORY_HOTPLUG
+choice
+ prompt "Memory Hotplug Default Online Type"
+ default MHP_DEFAULT_ONLINE_TYPE_OFFLINE
help
+ Default memory type for driver managed hotplug memory.
We should call it "hotplugged memory" consistently, which it is from a
pure core-mm perspective ("add memory").
"Driver managed" reminds too much about add_memory_driver_managed(),
which is only one case. Maybe just drop the "e.g., page tables" from the
examples below.
+
This option sets the default policy setting for memory hotplug
onlining policy (/sys/devices/system/memory/auto_online_blocks) which
determines what happens to newly added memory regions. Policy setting
can always be changed at runtime.
+
+ The default is 'offline'.
+
+ Select offline to defer onlining to drivers and user policy.
+ Select auto to let the kernel choose what zones to utilize.
+ Select online_kernel to generally allow kernel usage of this memory.
+ Select online_movable to generally disallow kernel usage of this memory.
+
+ Example kernel usage would be page structs and page tables.
+
See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
- Say Y here if you want all hot-plugged memory blocks to appear in
- 'online' state by default.
- Say N here if you want the default policy to keep all hot-plugged
- memory blocks in 'offline' state.
+config MHP_DEFAULT_ONLINE_TYPE_OFFLINE
+ bool "offline"
+ help
+ Driver managed memory will not be onlined by default.
"Hotplugged memory"
> + Choose this for systems with drivers and user policy that> +
handle onlining of hotplug memory policy.
> +> +config MHP_DEFAULT_ONLINE_TYPE_ONLINE_AUTO
+ bool "auto"
+ help
+ Select this if you want the kernel to automatically online
> + memory into the zone it thinks is reasonable. This memory
hotplugged memory
+ may be utilized for kernel data (e.g. page tables).
+
+config MHP_DEFAULT_ONLINE_TYPE_ONLINE_KERNEL
+ bool "kernel"
+ help
+ Select this if you want the kernel to automatically online
+ hotplug memory into a zone capable of being used for kernel
+ data (e.g. page tables). This typically means ZONE_NORMAL.
+
+config MHP_DEFAULT_ONLINE_TYPE_ONLINE_MOVABLE
+ bool "movable"
+ help
+ Select this if you want the kernel to automatically online
+ hotplug memory into ZONE_MOVABLE. This memory will generally
+ not be utilized for kernel data (e.g. page tables).
+
+ This should only be used when the admin knows sufficient
+ ZONE_NORMAL memory is available to describe hotplug memory,
+ otherwise hotplug memory may fail to online. For example,
+ sufficient kernel-capable memory (ZONE_NORMAL) must be
+ available to allocate page structs to describe ZONE_MOVABLE.
+
+endchoice
config MEMORY_HOTREMOVE
bool "Allow for memory hot remove"
diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
index 3b6f93962481..e3655f07dd6e 100644
--- a/mm/memory_hotplug.c
+++ b/mm/memory_hotplug.c
@@ -219,11 +219,30 @@ void put_online_mems(void)
bool movable_node_enabled = false;
-#ifndef CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE
-int mhp_default_online_type = MMOP_OFFLINE;
-#else
-int mhp_default_online_type = MMOP_ONLINE;
-#endif
+static int mhp_default_online_type = -1;
+int mhp_get_default_online_type(void)
+{
+ if (mhp_default_online_type >= 0)
+ return mhp_default_online_type;
+
+ if (IS_ENABLED(CONFIG_MHP_DEFAULT_ONLINE_TYPE_OFFLINE))
+ mhp_default_online_type = MMOP_OFFLINE;
+ else if (IS_ENABLED(CONFIG_MHP_DEFAULT_ONLINE_TYPE_ONLINE_AUTO))
+ mhp_default_online_type = MMOP_ONLINE;
+ else if (IS_ENABLED(CONFIG_MHP_DEFAULT_ONLINE_TYPE_ONLINE_KERNEL))
+ mhp_default_online_type = MMOP_ONLINE_KERNEL;
+ else if (IS_ENABLED(CONFIG_MHP_DEFAULT_ONLINE_TYPE_ONLINE_MOVABLE))
+ mhp_default_online_type = MMOP_ONLINE_MOVABLE;
+ else
+ mhp_default_online_type = MMOP_OFFLINE;
What would be nice is if we could use the symbols from Kconfig, to then
only do
mhp_default_online_type = CONFIG_MHP_DEFAULT_ONLINE_TYPE;
But as far as I know, that's not possible.
Thanks!
Acked-by: David Hildenbrand <david@xxxxxxxxxx>
--
Cheers,
David / dhildenb
