[Devel] VZ_QUOTA without VE

Konstantin G. Khlebnikov koct9i at gmail.com
Sun Aug 8 02:13:57 PDT 2010


This is not a problem, because compilation with CONFIG_VE=n deadly
broken (by design)

On Sat, Aug 7, 2010 at 2:01 AM, Kirill Kolyshkin <kolyshkin at gmail.com> wrote:
> Plaza reply @ devel list
>
> sent from my google nexus one android mobile phone
>
> ---------- Forwarded message ----------
> From: "Dennis Schridde" <devurandom at gmx.net>
> Date: Aug 7, 2010 12:59 AM
> Subject: [Devel] VZ_QUOTA without VE
> To: <devel at openvz.org>
>
> Hello!
>
> I am looking at the 2.6.32 branch:
> Is it intentional that e.g.
> VZ_QUOTA can be enabled even if VE is not?
>
> Particularly problematic seems
> this:
> struct proc_dir_entry *proc_vz_dir;
> struct proc_dir_entry
> *glob_proc_vz_dir;
>
> That's in kernel/ve/veowner.c (including
> initialisation), but the quota code makes use of these variables.
> So for
> VZ_QUOTA and VE to be truly independent, the initialisation (as well as the
> symbol itself) would have to move out of veowner.c into some "neutral" space
> (imo).
>
> --Dennis
>




More information about the Devel mailing list