Hi Jean-Pierre,
Thank you for the fast reply! :)
On Fri, 28 Mar 2008, Jean-Pierre ANDRE wrote:
> I have just set the timeout again (more precisely removed
> the timeout removal from 1.2310SB.4),
I couldn't compile the CVS. When I did an './autogen.sh' I got
autoreconf: running: automake --add-missing --copy --force-missing
configure.ac:334: required file `libntfs-3g/libntfs-3g.pc.in' not found
configure.ac:334: required file `libntfs-3g/libntfs-3g.script.so.in' not found
configure.ac:334: required file `src/ntfs-3g.probe.8.in' not found
configure.ac:334: required file `include/fuse-lite/Makefile.in' not found
configure.ac:334: required file `libfuse-lite/Makefile.in' not found
autoreconf: automake failed with exit status: 1
I think these two commands should solve it:
cvs add libntfs-3g/libntfs-3g.pc.in libntfs-3g/libntfs-3g.script.so.in
src/ntfs-3g.probe.8.in include/fuse-lite/Makefile.am
libfuse-lite/Makefile.am
cvs commit
> and run the following in a sh (this is important) :
>
> rm file0 file1 file2
> echo file > file0
> ln file0 file1
> ln file0 file2
> ls -l file*
> echo more >> file1
> ls -l file*
>
> This is the result :
>
> -rw-r--r-- 3 root root 5 2008-03-28 12:51 file0
> -rw-r--r-- 2 root root 5 2008-03-28 12:51 file1
> -rw-r--r-- 3 root root 5 2008-03-28 12:51 file2
> -rw-r--r-- 3 root root 5 2008-03-28 12:51 file0
> -rw-r--r-- 3 root root 10 2008-03-28 12:51 file1
> -rw-r--r-- 3 root root 5 2008-03-28 12:51 file2
>
> If I have not been mistaken, this shows the fuse cacheing
> still ignores the hard links, and considers the hard linked
> files as different files.
You are not mistaken :)
> This was the cause of three tests failing in the Posix test
> suite (which could be hidden by inserting delays so that
> attributes are flushed out of cache).
Could you please tell me the three specific test cases? I'll compare them
with my results.
I think you must be right. This issue was indeed suspended temporarily and
workarounded. I remember that Miklos wrote that hard links are not easy to
solve, and he doesn't feel it to be important enough. But having fuse-lite
now, I may take a quick look ...
> I have the timeout suppressed only on my own versions which
> have their own inode cacheing, so fuse not cacheing has no
> much effect on the overall performance (about 5%, with my
> own cacheing improving by 50%).
Yep. At the moment we need attr_timeout=0 for correctness and the caching
is more efficient in general.
Regards,
Szaka
> > Message du 28/03/08 12:43
> > De : "Szabolcs Szakacsits" <[EMAIL PROTECTED]>
> > A : "Jean-Pierre AndrĂŠ" <[EMAIL PROTECTED]>
> > Copie Ă : [email protected]
> > Objet : attr_timeout=0 (was: Re: Release candidate for ntfs-3g with file
> > ownership and permissions)
> >
> >
> > Hi Jean-Pierre,
> >
> > On Wed, 16 Jan 2008, [ISO-8859-1] Jean-Pierre AndrĂŠ wrote:
> > > Szabolcs Szakacsits wrote:
> > > >
> > > > Bonnie++ result:
> > > > File creation: 10% improvement
> > > > File stat: 50% decrease <-- disabled fuse cache, quite
> > > > probably
> > >
> > > You can just remove the FUSE timeout option to confirm your diagnosis.
> >
> > Yes, it was the attr_timeout=0.
> >
> > Do we __really__ need this? I believed that the FUSE kernel module in
> > FUSE 2.7.3 has fixed all attribute update problems we have found. The
> > FUSE module in the kernel is not up-to-date yet, afaik.
> >
> > I'm asking this because enabling more aggressive FUSE caching would
> > increase stat performance about 3000% and would help in overall too.
> >
> > Regards,
> > Szaka
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
ntfs-3g-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ntfs-3g-devel
