Thread (15 messages) 15 messages, 4 authors, 2016-09-09

Re: [PATCH v4 RESEND 1/2] thp, dax: add thp_get_unmapped_area for pmd mappings

From: Kani, Toshimitsu <hidden>
Date: 2016-08-29 20:44:57
Also in: linux-fsdevel, lkml, nvdimm

On Mon, 2016-08-29 at 12:34 -0700, Dan Williams wrote:
On Mon, Aug 29, 2016 at 12:11 PM, Toshi Kani [off-list ref]
wrote:
quoted
When CONFIG_FS_DAX_PMD is set, DAX supports mmap() using pmd page
size.  This feature relies on both mmap virtual address and FS
block (i.e. physical address) to be aligned by the pmd page size.
Users can use mkfs options to specify FS to align block
allocations. However, aligning mmap address requires code changes
to existing applications for providing a pmd-aligned address to
mmap().

For instance, fio with "ioengine=mmap" performs I/Os with mmap()
[1]. It calls mmap() with a NULL address, which needs to be changed
to provide a pmd-aligned address for testing with DAX pmd mappings.
Changing all applications that call mmap() with NULL is
undesirable.

Add thp_get_unmapped_area(), which can be called by filesystem's
get_unmapped_area to align an mmap address by the pmd size for
a DAX file.  It calls the default handler, mm->get_unmapped_area(),
to find a range and then aligns it for a DAX file.

The patch is based on Matthew Wilcox's change that allows adding
support of the pud page size easily.
 :
Reviewed-by: Dan Williams <redacted>
Great!
...with one minor nit:

quoted
 include/linux/huge_mm.h |    7 +++++++
 mm/huge_memory.c        |   43
+++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 50 insertions(+)
diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
index 6f14de4..4fca526 100644
--- a/include/linux/huge_mm.h
+++ b/include/linux/huge_mm.h
@@ -87,6 +87,10 @@ extern bool is_vma_temporary_stack(struct
vm_area_struct *vma);

 extern unsigned long transparent_hugepage_flags;

+extern unsigned long thp_get_unmapped_area(struct file *filp,
+               unsigned long addr, unsigned long len, unsigned
long pgoff,
+               unsigned long flags);
+
 extern void prep_transhuge_page(struct page *page);
 extern void free_transhuge_page(struct page *page);
@@ -169,6 +173,9 @@ void put_huge_zero_page(void);
 static inline void prep_transhuge_page(struct page *page) {}

 #define transparent_hugepage_flags 0UL
+
+#define thp_get_unmapped_area  NULL
Lets make this:

static inline unsigned long thp_get_unmapped_area(struct file *filp,
               unsigned long addr, unsigned long len, unsigned long
pgoff,
               unsigned long flags)
{
    return 0;
}

...to get some type checking in the CONFIG_TRANSPARENT_HUGEPAGE=n
case.
Per get_unmapped_area() in mm/mmap.c, I think we need to set it to NULL
when we do not override current->mm->get_unmapped_area.

Thanks!
-Toshi

�{.n�+���zwZ��,j���n�˛���m�b��f�)���w+h����&�K�rJ+�Z+��ފw�r��L2Ǟ��i��0�X�����?��zf���ky���[hv�������zf��K�rJ+��
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help