[Gluster-devel] [PATCH 3/3 v3] vfs_glusterfs: Samba VFS module for glusterfs

Christopher R. Hertel crh at ubiqx.mn.org
Thu May 23 14:27:14 UTC 2013


On Mon, May 20, 2013 at 04:10:32PM -0700, Anand Avati wrote:
> On 5/20/13 3:38 PM, Andrew Bartlett wrote:
> >On Mon, 2013-05-20 at 03:11 -0400, Anand Avati wrote:
> >>Implement a Samba VFS plugin for glusterfs based on gluster's gfapi.
> >>This is a "bottom" vfs plugin (not something to be stacked on top of
> >>another module), and translates (most) calls into closest actions
> >>on gfapi.
> >
> >Thanks for getting back to us.
> >
> >The comment I have is however still around the realloc stuff.  Is
> >talloc_realloc() so slow that there is such value in having the code
> >realloc 1024 or 16 more entries than are required?
> >
> >It would seem clearer if we just allocated exactly the array size
> >needed.   That would also allow bounds checkers like valgrind to
> >correctly know the 'real' allocated size.
> >
> >It just smells to me like premature optimisation, and means you have to
> >do the partial array memset() which always makes me nervous.
> >
> >That said, this isn't an objection, just an observation.
> >
> 
> It wasn't really about talloc_XXX() being slow. It is just a standard 
> technique generally used in many places. I can change the code to 
> increase allocation count by one if anybody else too shares your opinion.

"Standard techniques" are, in general, only standard within the scope of 
one's experience.  The Samba Team has, over the course of 20+ years, 
developed many of their own standards.  What Abartlet is describing is an 
approach common to Samba.

Since the module is to be included in the Samba distribution, I'd vote 
in favor of using Samba conventions.  That will help us get the code 
through the review process and into the Master branch.

Chris -)-----




More information about the Gluster-devel mailing list