Re: [RFC] zram: support page-based parallel write
From: Minchan Kim
Date: Wed Sep 07 2016 - 22:31:48 EST
Hello,
On Tue, Sep 06, 2016 at 10:22:20AM +0200, Andreas Mohr wrote:
> On Tue, Sep 06, 2016 at 04:24:17PM +0900, Minchan Kim wrote:
> > @@ -1464,6 +1908,9 @@ static int __init zram_init(void)
> > num_devices--;
> > }
> >
> > + if (create_workers())
> > + goto out_error;
> > +
> > return 0;
> >
> > out_error:
> > @@ -1474,6 +1921,7 @@ static int __init zram_init(void)
> > static void __exit zram_exit(void)
> > {
> > destroy_devices();
> > + destroy_workers();
> > }
>
>
> Asymmetry --> "BUG".
>
> ...right?
destory_workers checks workers list so if it's empty, it doesn
nothing. Anyway, I am chaning thread management model now so
it should be changed, too. :)
>
> (I have to admit that current implementation structure
> is not easy to follow,
> thus I'm not fully sure)
>
>
> Thanks for working in this important area!
Thanks for the interest.