[PATCH] dax: allowing MAP_SYNC for device-dax mmap() to succeed

[PATCH] dax: allowing MAP_SYNC for device-dax mmap() to succeed

[Dave Jiang]

MAP_SYNC is basically a nop for device-dax. However allowing the MAP_SYNC
flag to succeed on device-dax would make it consistent with fs-dax and
reduces confusion for the application writer. This allows the application to
assume that it does not need to call fsync() after writes to device-dax
mappings.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
---
 drivers/dax/device.c |    2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/dax/device.c b/drivers/dax/device.c
index 37be5a306c8f..374b6718f6c5 100644
--- a/drivers/dax/device.c
+++ b/drivers/dax/device.c
@@ -19,6 +19,7 @@
 #include <linux/dax.h>
 #include <linux/fs.h>
 #include <linux/mm.h>
+#include <linux/mman.h>
 #include "dax-private.h"
 #include "dax.h"
 
@@ -530,6 +531,7 @@ static const struct file_operations dax_fops = {
 	.release = dax_release,
 	.get_unmapped_area = dax_get_unmapped_area,
 	.mmap = dax_mmap,
+	.mmap_supported_flags = MAP_SYNC,
 };
 
 static void dev_dax_release(struct device *dev)

_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

Re: [PATCH] dax: allowing MAP_SYNC for device-dax mmap() to succeed

[Dan Williams]

I'll change the "allowing" to "allow" in the subject so we don't have
a gerund phrase.

On Thu, Apr 19, 2018 at 11:22 AM, Dave Jiang <dave.jiang@intel.com> wrote:
> MAP_SYNC is basically a nop for device-dax. However allowing the MAP_SYNC

Not "bascially", it *is* a nop for device-dax. I'll change this second
sentence to start "Allow MAP_SYNC to succeed on device-dax to
eliminate special casing between device-dax and fs-dax as to when the
flag can be specified."

> flag to succeed on device-dax would make it consistent with fs-dax and
> reduces confusion for the application writer. This allows the application to
> assume that it does not need to call fsync() after writes to device-dax
> mappings.

I'll change this last sentence to say "Device-dax users already
implicitly assume that they do not need to call fsync(), enable them
to explicitly check for this capability".
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm