On Mon, 2008-07-07 at 19:24 +0900, Yasunori Goto wrote:

quoted

include/linux/mempolicy.h already has a !NUMA section could we not just
define policy_zone as 0 in that and leave this code unconditionally
compiled?  Perhaps also adding a NUMA_BUILD && to this 'if' should that
be clearer.

Ah, yes. It's better. :-)

ok, the new patch below defines policy_zone as 0 in the !NUMA section. The
compiler will automatically omit the if statement w/o NUMA in this case.

quoted

But this does make me feel uneasy.  Are we really saying all memory on
an s390 is migratable.  That seems unlikely. As I understand the NUMA
case, we only allow migration of memory in the last zone (last two if we
have a MOVABLE zone) why are things different just because we have a
single 'node'.  Hmmm.  I suspect strongly that something is missnamed
more than there is a problem.

If my understanding is correct, even if this policy_zone check is removed,
page isolation will just fail due to some busy pages.
In hotplug case, it means giving up of hotremoving,
and kernel must be rollback to make same condition of previous
starting offline_pages().
This check means just "early" check, but not so effective for hotremoving,
I think....

It seems to me that this policy_zone check in vma_migratable() is not
called at all for the offline_pages() case, only for some NUMA system calls
that we don't support on s390. As Yasunori Goto said, page isolation checks
should do the job for memory hotremove via offline_pages(), independent from
any policy_zone setting. Any more thoughts on this?

Thanks,
Gerald
---

Subject: [PATCH] Make CONFIG_MIGRATION available for s390

From: Gerald Schaefer <redacted>

We'd like to support CONFIG_MEMORY_HOTREMOVE on s390, which depends on
CONFIG_MIGRATION. So far, CONFIG_MIGRATION is only available with NUMA
support.

This patch makes CONFIG_MIGRATION selectable for architectures that define
ARCH_ENABLE_MEMORY_HOTREMOVE. When MIGRATION is enabled w/o NUMA, the kernel
won't compile because of a missing migrate() function in vm_operations_struct
and a missing policy_zone reference in vma_migratable(). To avoid this,
policy_zone is defined as 0 for !NUMA, and the vm_ops migrate() definition
is moved from '#ifdef CONFIG_NUMA' to '#ifdef CONFIG_MIGRATION'.

Signed-off-by: Gerald Schaefer <redacted>
---

 include/linux/mempolicy.h |    1 +
 include/linux/mm.h        |    2 ++
 mm/Kconfig                |    2 +-
 3 files changed, 4 insertions(+), 1 deletion(-)

Index: linux-2.6/include/linux/mm.h
===================================================================

--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h

@@ -193,6 +193,8 @@ struct vm_operations_struct {
 	 */
 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 					unsigned long addr);
+#endif
+#ifdef CONFIG_MIGRATION
 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 		const nodemask_t *to, unsigned long flags);
 #endif

Index: linux-2.6/mm/Kconfig
===================================================================

--- linux-2.6.orig/mm/Kconfig
+++ linux-2.6/mm/Kconfig

@@ -174,7 +174,7 @@ config SPLIT_PTLOCK_CPUS
 config MIGRATION
 	bool "Page migration"
 	def_bool y
-	depends on NUMA
+	depends on NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE
 	help
 	  Allows the migration of the physical location of pages of processes
 	  while the virtual addresses are not changed. This is useful for

Index: linux-2.6/include/linux/mempolicy.h
===================================================================

--- linux-2.6.orig/include/linux/mempolicy.h
+++ linux-2.6/include/linux/mempolicy.h

@@ -222,6 +222,7 @@ extern int mpol_to_str(char *buffer, int
 #endif
 #else
 
+#define policy_zone	0
 struct mempolicy {};
 
 static inline int mpol_equal(struct mempolicy *a, struct mempolicy *b)


--

To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>