On Fri, 11 Nov 2016 20:21:41 -0800 Dan Williams <dan.j.willi...@intel.com>
wrote:
> Mark dax vmas as not migratable to exclude them from task_numa_work().
> This is especially relevant for device-dax which wants to ensure
> predictable access latency and not incur periodic faults.
>
> ...
>
> @@ -177,6 +178,9 @@ static inline bool vma_migratable(struct vm_area_struct
> *vma)
> if (vma->vm_flags & (VM_IO | VM_PFNMAP))
> return false;
>
> + if (vma_is_dax(vma))
> + return false;
> +
> #ifndef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
> if (vma->vm_flags & VM_HUGETLB)
> return false;
I don't think the reader could figure out why this code is here, so... this?
---
a/include/linux/mempolicy.h~mm-disable-numa-migration-faults-for-dax-vmas-fix
+++ a/include/linux/mempolicy.h
@@ -180,6 +180,10 @@ static inline bool vma_migratable(struct
if (vma->vm_flags & (VM_IO | VM_PFNMAP))
return false;
+ /*
+ * DAX device mappings require predictable access latency, so avoid
+ * incurring periodic faults.
+ */
if (vma_is_dax(vma))
return false;
_
