Re: freevxfs: hp-ux support. patchset 1-7/7 rev 2

["Krzysztof Błaszkowski" <kb@sysmikro.com.pl>]

On Thu, 2016-06-02 at 01:25 -0700, Christoph Hellwig wrote:
> On Wed, Jun 01, 2016 at 11:23:32AM +0200, Krzysztof B??aszkowski wrote:
> > I think that there are no "kfree(pfp); kfree(sfp); return 0;" in
> > vxfs_read_fshead() still. Are pfp and sfp needed anywhere ? I am sure
> > they are not so there is a memory leak without these kfrees every mount.
> > 
> > I am not sure absolutely of that read_fshead() is missing these kfrees
> > because I have seen just these diffs, anyway I did not notice "+kfree".
> 
> The frees are still missing.  Do you want to send me a patch for those?

I see. Don't hesitate to add them, I was thinking that creating a
special patch because of these just two kfree()s is just too much formal
work.

> 
> > needles to say that I prefer to have limited scope of visibility of
> > inode_cachep to the inode.c only.
> 
> In my tree the visibility is in vxfs_super.c only.  Given that the
> alloc/destroy methods are super operations that seems to fit better
> as they can have local scope, too.

I see. We differ in point of views. I presumed that it is more
consistent to have everything regarding inodes creation
(getting)/destroying, etc in the inode.c. Anyway this is just pure
academic debate, we could argue on anything else (weather, politics and
politicians - hot topic especially in Poland) as well.

You decide. You enjoy more your approach - good enough.


-- 
Krzysztof Blaszkowski