Date: Mon, 15 Jun 2015 11:28:38 +0200
From: Petr Mladek <pmladek@suse.cz>
To: Tejun Heo <tj@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>, Oleg Nesterov <oleg@redhat.com>,
        Ingo Molnar <mingo@redhat.com>, Peter Zijlstra <peterz@infradead.org>,
        Richard Weinberger <richard@nod.at>,
        Steven Rostedt <rostedt@goodmis.org>,
        David Woodhouse <dwmw2@infradead.org>, linux-mtd@lists.infradead.org,
        Trond Myklebust <trond.myklebust@primarydata.com>,
        Anna Schumaker <anna.schumaker@netapp.com>, linux-nfs@vger.kernel.org,
        Chris Mason <clm@fb.com>,
        "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Jiri Kosina <jkosina@suse.cz>, Borislav Petkov <bp@suse.de>,
        Michal Hocko <mhocko@suse.cz>, live-patching@vger.kernel.org,
        linux-api@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH 07/18] kthread: Make iterant kthreads freezable by
 default
Message-ID: <20150615092838.GK9409@pathway.suse.cz>
References: <1433516477-5153-1-git-send-email-pmladek@suse.cz>
 <1433516477-5153-8-git-send-email-pmladek@suse.cz>
 <20150609072003.GY21465@mtj.duckdns.org>
 <20150609155313.GC9409@pathway.suse.cz>
 <20150610043154.GG11955@mtj.duckdns.org>
 <20150612132440.GH9409@pathway.suse.cz>
 <20150613232222.GB346@mtj.duckdns.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <20150613232222.GB346@mtj.duckdns.org>
Sender: linux-nfs-owner@vger.kernel.org

On Sat 2015-06-13 18:22:22, Tejun Heo wrote:
> > I try to better understand why freezer is considered to be a blunt
> > tool. Is it because it is a generic API, try_to_freeze() is put on
> > "random" locations, so that it does not define the safe point
> > precisely enough?
>
> Not that.  I don't know how to explain it better.  Hmmm... okay, let's
> say there's a shared queue Q and N users o fit.  If you wanna make Q
> empty and keep it that way for a while, the right thing to do is
> blocking new queueing and then wait till Q drains - you choke the
> entity that you wanna control.
>
> Instead of that, freezer is trying to block the "N" users part.  In
> majority of cases, it blocks enough but it's pretty difficult to be
> sure whether you actually got all N of them (as some of them may not
> involve kthreads at all or unfreezable kthreads might end up doing
> those operations too on corner cases) and it's also not that clear
> whether blocking the N users actually make Q empty.  Maybe there are
> things which can be in flight asynchronously on Q even all its N users
> are blocked.  This is inherently finicky.

I feel convinced that it does not make sense to make kthreads
freezable by default and that we should not use it when not
necessary.

Thanks a lot for patience and so detailed explanation.

Best Regards,
Petr