Patchset to add asynchronous open/close to master

[Volker Lendecke]

On Tue, Jun 19, 2012 at 03:16:52PM -0700, Jeremy Allison wrote:
> On Thu, Jun 14, 2012 at 05:24:42PM -0700, Jeremy Allison wrote:
> > 
> > Actually, thinking about this as both Simo and Metze
> > have complained, I will remove any references to SMB_VFS_NEXT*
> > calls in the code, and call directly to the underlying
> > syscalls/and or smbd code. Looking at this again there
> > isn't a lot of duplication anyway, so I'm not gaining
> > much.
> > 
> > It is confusing, and is only a lazy way of doing
> > the right thing - so I'll just do the right thing :-).
> > 
> > Thanks Metze + Simo :-).
> 
> So here is the updated patchset for master.

In a commit you mentioned that it has a race condition
fixed. Can you explain that? I am very worried that we put
an open potentially creating a file into a thread and (as it
seems to me) we fix up potential screwups later with a
fchown. This to me far too badly smells like a potential
race condition. The only way I see to correctly do this is
by a helper process that does our normal become_user and not
switch back to become_root(). Threads for my taste are too
dangerous for this.

With best regards,

Volker Lendecke

-- 
SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
http://www.sernet.de, mailto:kontakt at sernet.de