Skip to contentSkip to conversation

mm: Move change_prot_numa outside CONFIG_ARCH_USES_NUMA_PROT_NONE

PATCH In review Aneesh Kumar K.V 3 participants 3 messages Dec 5, 2013 – Dec 6, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm: Move change_prot_numa outside CONFIG_ARCH_USES_NUMA_PROT_NONE

From: "Aneesh Kumar K.V" <[email protected]>

change_prot_numa should work even if _PAGE_NUMA != _PAGE_PROTNONE. On archs like ppc64 that don't use _PAGE_PROTNONE and also have a separate page table outside linux pagetable, we just need to make sure that when calling change_prot_numa we flush the hardware page table entry so that next page access result in a numa fault.

We still need to make sure we use the numa faulting logic only when CONFIG_NUMA_BALANCING is set. This implies the migrate-on-fault (Lazy migration) via mbind will only work if CONFIG_NUMA_BALANCING is set.

---
Previous discussion around the patch can be found at
http://article.gmane.org/gmane.linux.kernel.mm/109305

changes from V2:
* Move the numa faulting definition within CONFIG_NUMA_BALANCING

 include/linux/mm.h | 2 +-
 mm/mempolicy.c     | 5 ++---
 2 files changed, 3 insertions(+), 4 deletions(-)
include/linux/mm.h +1 -1
index 1cedd000cf29..a7b4e310bf42 100644
@@ -1842,7 +1842,7 @@ static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
 }
 #endif
 
-#ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
+#ifdef CONFIG_NUMA_BALANCING
 unsigned long change_prot_numa(struct vm_area_struct *vma,
 			unsigned long start, unsigned long end);
 #endif
mm/mempolicy.c +2 -3
index eca4a3129129..9f73b29d304d 100644
@@ -613,7 +613,7 @@ static inline int queue_pages_pgd_range(struct vm_area_struct *vma,
 	return 0;
 }
 
-#ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
+#ifdef CONFIG_NUMA_BALANCING
 /*
  * This is used to mark a range of virtual addresses to be inaccessible.
  * These are later cleared by a NUMA hinting fault. Depending on these
@@ -627,7 +627,6 @@ unsigned long change_prot_numa(struct vm_area_struct *vma,
 			unsigned long addr, unsigned long end)
 {
 	int nr_updated;
-	BUILD_BUG_ON(_PAGE_NUMA != _PAGE_PROTNONE);
 
 	nr_updated = change_protection(vma, addr, end, vma->vm_page_prot, 0, 1);
 	if (nr_updated)
@@ -641,7 +640,7 @@ static unsigned long change_prot_numa(struct vm_area_struct *vma,
 {
 	return 0;
 }
-#endif /* CONFIG_ARCH_USES_NUMA_PROT_NONE */
+#endif /* CONFIG_NUMA_BALANCING */
 
 /*
  * Walk through page tables and collect pages to be migrated.
Signature · 1 line
1.8.3.2
Reviewed-by on v1 1/1
#1 lore raw
Quoted 9 lines · Aneesh Kumar K.V

From: "Aneesh Kumar K.V" <[email protected]>

change_prot_numa should work even if _PAGE_NUMA != _PAGE_PROTNONE. On archs like ppc64 that don't use _PAGE_PROTNONE and also have a separate page table outside linux pagetable, we just need to make sure that when calling change_prot_numa we flush the hardware page table entry so that next page access result in a numa fault.

Quoted 6 lines

We still need to make sure we use the numa faulting logic only when CONFIG_NUMA_BALANCING is set. This implies the migrate-on-fault (Lazy migration) via mbind will only work if CONFIG_NUMA_BALANCING is set.

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

Acked-by on v1 1/1
#2 lore raw
In reply to Aneesh Kumar K.V
Quoted 9 lines · Aneesh Kumar K.V

From: "Aneesh Kumar K.V" <[email protected]>

change_prot_numa should work even if _PAGE_NUMA != _PAGE_PROTNONE. On archs like ppc64 that don't use _PAGE_PROTNONE and also have a separate page table outside linux pagetable, we just need to make sure that when calling change_prot_numa we flush the hardware page table entry so that next page access result in a numa fault.

Quoted 6 lines

We still need to make sure we use the numa faulting logic only when CONFIG_NUMA_BALANCING is set. This implies the migrate-on-fault (Lazy migration) via mbind will only work if CONFIG_NUMA_BALANCING is set.

You're right on that there is no direct dependance on numa balancing and use of prot_none. The BUILD_BUG_ON was to flag very clearly that arches wanting to support automatic NUMA balancing must ensure such things as

o _PAGE_NUMA is defined
o setting _PAGE_NUMA traps a fault and the fault can be uniquely identified as being a numa hinting fault
o that pte_present still returns true for pte_numa pages even though the underlying present bit may be cleared. Otherwise operations like following and copying ptes will get confused
o shortly, arches will also need to avoid taking references on pte_numa pages in get_user_pages to account for hinting faults properly

I guess the _PAGE_NUMA parts will already be caught by other checks and the rest will fall out during testing so it's ok to remove.

Signature · 8 lines
Mel Gorman
SUSE Labs

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