[PATCH] enable possibility of ALL different atime mount options

[PATCH] enable possibility of ALL different atime mount options

[Roland Eggner]

Utako Kusaka has provided this patch
http://oss.sgi.com/archives/xfs/2007-10/msg00168.html
It works for me GRACEFULLY with ALL different atime mount options: atime|noatime|relatime.
Note, the mount options are effective, not only accepted.
Prior to this patch, mount option atime has been ignored.

Please put the patch to mainline, it would save worries for users, who need the traditional atime
bahavior.  Think of some corner cases for desktop usage.

-- 
Roland Eggner

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] enable possibility of ALL different atime mount options

[Eric Sandeen]

Roland Eggner wrote:
> Utako Kusaka has provided this patch
> http://oss.sgi.com/archives/xfs/2007-10/msg00168.html
> It works for me GRACEFULLY with ALL different atime mount options: atime|noatime|relatime.
> Note, the mount options are effective, not only accepted.
> Prior to this patch, mount option atime has been ignored.



> Please put the patch to mainline, it would save worries for users, who need the traditional atime
> bahavior.  Think of some corner cases for desktop usage.
> 

Which specific problem is this fixing for you?

you say that atime is ignored; but:

default mounts:

[root@inode test]# echo foo > atimetestfile
[root@inode test]# stat atimetestfile
Access: 2009-01-30 04:36:47.470348714 -0600
[root@inode test]# cat atimetestfile
foo
[root@inode test]# stat atimetestfile
Access: 2009-01-30 04:36:57.450339777 -0600

remount as noatime:

[root@inode test]# mount -o remount,noatime .
[root@inode test]# stat atimetestfile
Access: 2009-01-30 04:36:57.450339777 -0600
[root@inode test]# cat atimetestfile
foo
[root@inode test]# stat atimetestfile
Access: 2009-01-30 04:36:57.450339777 -0600

it would appear to be working...

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] enable possibility of ALL different atime mount options

[Eric Sandeen]

Eric Sandeen wrote:
> Roland Eggner wrote:
>> Utako Kusaka has provided this patch
>> http://oss.sgi.com/archives/xfs/2007-10/msg00168.html
>> It works for me GRACEFULLY with ALL different atime mount options: atime|noatime|relatime.
>> Note, the mount options are effective, not only accepted.
>> Prior to this patch, mount option atime has been ignored.
> 
> 
> 
>> Please put the patch to mainline, it would save worries for users, who need the traditional atime
>> bahavior.  Think of some corner cases for desktop usage.
>>
> 
> Which specific problem is this fixing for you?
> 
> you say that atime is ignored; but:

Sorry, I probably replied too quickly.  But I am still curious to know
what problem you are seeing that this patch fixes.

Thanks,
-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] enable possibility of ALL different atime mount options

[Roland Eggner]

On Friday Jan 30th 2009 at 05:47:34 Eric Sandeen wrote:
> Eric Sandeen wrote:
> > Roland Eggner wrote:
> >> Utako Kusaka has provided this patch
> >> http://oss.sgi.com/archives/xfs/2007-10/msg00168.html
> >> It works for me GRACEFULLY with ALL different atime mount options: atime|noatime|relatime.
> >> Note, the mount options are effective, not only accepted.
> >> Prior to this patch, mount option atime has been ignored.
> > 
> > 
> > 
> >> Please put the patch to mainline, it would save worries for users, who need the traditional atime
> >> bahavior.  Think of some corner cases for desktop usage.
> >>
> > 
> > Which specific problem is this fixing for you?
> > 
> > you say that atime is ignored; but:
> 
> Sorry, I probably replied too quickly.  But I am still curious to know
> what problem you are seeing that this patch fixes.
> 
> Thanks,
> -Eric

On Wed Oct 17th 2007 at 13:56:00 +0900 Utako Kusaka wrote:
>The atime is changed for reading but it returns to a previous value
>after unmount.
And his patch fixes this, nothing more and nothing less.
Think of desktop machines, where you sometimes need precise atime stamps,
and where atime only inode updates don't hurt performance because CPUs are
almost always idle.  Think of e.g. important cron jobs, you want to know
exactly, if at all, when and in which sequence they were executed.

-- 
Roland Eggner

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs