Memory hotplug for new nodes v.4.(rebuild zonelists after online pages)
PATCH 12/17 Memory hotplug for new nodes v.4.(rebuild zonelists after online pages)
In current code, zonelist is considered to be build once, no modification. But MemoryHotplug can add new zone/pgdat. It must be updated.
This patch modifies build_all_zonelists().
By this, build_all_zonelist() can reconfig pgdat's zonelists.
To update them safety, this patch use stop_machine_run(). Other cpus don't touch among updating them by using it.
In previous version (V2), kernel updated them after zone initialization. But present_page of its new zone is still 0, because online_page() is not called yet at this time.Build_zonelists() checks present_pages to find present zone. It was too early. So, I changed it after online_pages().
mm/memory_hotplug.c | 12 ++++++++++++ mm/page_alloc.c | 26 +++++++++++++++++++++----- 2 files changed, 33 insertions(+), 5 deletions(-)
Index: pgdat8/mm/page_alloc.c
=================================--- pgdat8.orig/mm/page_alloc.c 2006-03-17 13:53:39.194026730 +0900
+++ pgdat8/mm/page_alloc.c 2006-03-17 13:53:45.530940715 +0900
@@ -37,6 +37,7 @@
#include <linux/nodemask.h>
#include <linux/vmalloc.h>
#include <linux/mempolicy.h>
+#include <linux/stop_machine.h>
#include <asm/tlbflush.h>
#include "internal.h"
@@ -1765,14 +1766,29 @@ static void __init build_zonelists(pg_da
#endif /* CONFIG_NUMA */
-void __init build_all_zonelists(void)
+/* return values int ....just for stop_machine_run() */ +static int __meminit __build_all_zonelists(void *dummy)
{
- int i;
+ int nid;
+ for_each_online_node(nid)
+ build_zonelists(NODE_DATA(nid));
+ return 0;
+}
+
+void __meminit build_all_zonelists(void)
+{
+ if (system_state = SYSTEM_BOOTING) {
+ __build_all_zonelists(0);
+ cpuset_init_current_mems_allowed();
+ } else {
+ /* we have to stop all cpus to guaranntee there is no user
+ of zonelist */
+ stop_machine_run(__build_all_zonelists, NULL, NR_CPUS);
+ /* cpuset refresh routine should be here */
+ }
- for_each_online_node(i)
- build_zonelists(NODE_DATA(i));
printk("Built %i zonelists\n", num_online_nodes());
- cpuset_init_current_mems_allowed();
+
}
/*
Index: pgdat8/mm/memory_hotplug.c
=================================--- pgdat8.orig/mm/memory_hotplug.c 2006-03-17 13:53:38.274104866 +0900
+++ pgdat8/mm/memory_hotplug.c 2006-03-17 13:53:40.712581399 +0900
@@ -123,6 +123,7 @@ int online_pages(unsigned long pfn, unsi
unsigned long flags;
unsigned long onlined_pages = 0;
struct zone *zone;
+ int need_refresh_zonelist = 0;
/*
* This doesn't need a lock to do pfn_to_page().
@@ -135,6 +136,14 @@ int online_pages(unsigned long pfn, unsi
grow_pgdat_span(zone->zone_pgdat, pfn, pfn + nr_pages); pgdat_resize_unlock(zone->zone_pgdat, &flags);
+ /*
+ * If this zone is not populated, then it is not in zonelist.
+ * This means the page allocator ignores this zone.
+ * So, zonelist must be updated after online.
+ */
+ if (!populated_zone(zone))
+ need_refresh_zonelist = 1;
+
for (i = 0; i < nr_pages; i++) {
struct page *page = pfn_to_page(pfn + i); online_page(page);
@@ -145,6 +154,9 @@ int online_pages(unsigned long pfn, unsi
setup_per_zone_pages_min();
+ if (need_refresh_zonelist)
+ build_all_zonelists();
+
return 0;
}
Signature · 1 line
Yasunori Goto
Review int online_pages(unsigned long pfn, unsi · line 126
Quoted 6 lines · Yasunori Goto
+++
pgdat8/mm/memory_hotplug.c2006-03-17 13:53:40.712581399 +0900(quoted diff) +1
-0@@ -123,6 +123,7 @@ int online_pages(unsigned long pfn, unsi unsigned long flags; unsigned long onlined_pages = 0; struct zone *zone; + int need_refresh_zonelist = 0;
I'd make this "need_to_rebuild_zonelists" or "need_zonelists_rebuild". I think those sound a little bit better.
Plus, it makes even more sense when you see:
if (need_to_rebuild_zonelists)
build_all_zonelists();if the names all match up.
-- Dave