[Virtio-fs] Question: atime update behavior of virtiofs

[Masayoshi Mizuma]

Hi Vivek,

On Mon, Dec 16, 2019 at 09:54:58AM -0500, Vivek Goyal wrote:
> On Fri, Dec 13, 2019 at 12:57:33PM -0500, Masayoshi Mizuma wrote:
> > Hello,
> > 
> > Let me ask questions about atime update behavior of virtiofs.
> > 
> > Background:
> > 
> > I believe the atime update behavior of virtiofs is same as
> > the underlying filesystem and the atime mount option of virtiofs
> > is ignored.
> 
> [ CC Miklos. He does not seem to be on virtio-fs list yet ]
> 
> Hi Masayoshi,
> 
> I have not paid too much of attention to atime behavior but I think it
> will be a combination of atime settings on virtio-fs as well as underlying
> filesystem.
> 
> In this particular case read request will go to virtiofsd server
> (especially in case of cache=none) and when virtiofsd access file
> on host, atime will be updated according to the atime settings on
> host filesystem.
> 
> 
> > 
> > For example, the underlying fileysystem is XFS and the atime option
> > is strictatime Then, mount the filesystem from the guest with noatime.
> > 
> > # mount -t virtiofs -o noatime myfs0 /mnt/myfs0
> > 
> > # cat /proc/mounts
> > ...
> > myfs0 /mnt/myfs0 virtiofs rw,noatime 0 0
> > ...
> > 
> > Check the atime as following script atime.h:
> > ============
> > #!/bin/bash
> > 
> > FILE=/mnt/myfs0/foo
> > 
> > rm -f $FILE
> > 
> > echo "hello" > $FILE
> > stat --format=%x $FILE
> > 
> > sleep 10
> > 
> > cat $FILE > /dev/null
> > stat --format=%x $FILE
> > ============
> > 
> > I can see the following result. The atime is updated even if the mount
> > option is noatime.
> > 
> > ~]# ./atime.sh
> > 2019-12-13 12:32:46.733024358 -0500
> > 2019-12-13 12:32:56.752898507 -0500
> > ~]#
> > 
> > Questions:
> > 
> > Q1. Is the atime behavior a specification for virtiofs, that means
> >     it's not a bug?
> > 
> 
> I will not call it a bug. We have not actually defined atime behavior of
> virtio-fs as such and its dependency on host filesystem.

Got it, thanks.

> 
> > Q2. If Q1 is yes, I think it would be better if the atime behavior
> >     could be described to documents so that user can know the behavior.
> >     What is the appropriate document?
> >     - man page of virtiofsd which Dave is trying to merge?
> >       https://lists.gnu.org/archive/html/qemu-devel/2019-12/msg02475.html
> >     - man page of mount(8)?
> >     - The official page (https://virtio-fs.gitlab.io/)?
> >     - Or...?
> 
> It probably is a good idea to atleast add a section in
> Documentation/filesystem/virtiofs.rst and talk about atime behavior in
> different configurations.

I created a patch for virtiofs.rst to add the mount options and the
atime behavior as following.
Does it make sense?

From: Masayoshi Mizuma <m.mizuma at jp.fujitsu.com>
Date: Tue, 17 Dec 2019 15:55:20 -0500
Subject: [PATCH] virtio-fs: Add mount option and atime behavior to the doc

Add a section to show the mount option and a subsection to show
the atime behavior.

Signed-off-by: Masayoshi Mizuma <m.mizuma at jp.fujitsu.com>
---
 Documentation/filesystems/virtiofs.rst | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/Documentation/filesystems/virtiofs.rst b/Documentation/filesystems/virtiofs.rst
index 4f338e3cb3f7..8b04160061fd 100644
--- a/Documentation/filesystems/virtiofs.rst
+++ b/Documentation/filesystems/virtiofs.rst
@@ -37,6 +37,20 @@ Mount file system with tag ``myfs`` on ``/mnt``:
 Please see https://virtio-fs.gitlab.io/ for details on how to configure QEMU
 and the virtiofsd daemon.
 
+Mount options
+-------------
+
+virtiofs supports general VFS mount options, for example, remount,
+ro, rw, context, etc. It also supports FUSE mount options.
+
+atime behavior
+^^^^^^^^^^^^^^
+
+The atime-related mount options, for example, noatime, strictatime,
+are ignored. The atime behavior for virtiofs is the same as the
+underlying filesystem of the directory that has been exported
+on the host.
+
 Internals
 =========
 Since the virtio-fs device uses the FUSE protocol for file system requests, the
-- 
2.18.1