From: Josef Bacik Subject: Re: [PATCH 1/2] add hrtimer_sleep_ns helper function Date: Tue, 19 Aug 2008 15:39:47 -0400 Message-ID: <20080819193947.GC21749@unused.rdu.redhat.com> References: <20080806190819.GH27394@unused.rdu.redhat.com> <20080819191508.GD8318@parisc-linux.org> <20080819192215.GB21749@unused.rdu.redhat.com> <20080819193657.GE8318@parisc-linux.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Josef Bacik , linux-kernel@vger.kernel.org, rwheeler@redhat.com, tglx@linutronix.de, linux-fsdevel@vger.kernel.org, chris.mason@oracle.com, linux-ext4@vger.kernel.org To: Matthew Wilcox Return-path: Content-Disposition: inline In-Reply-To: <20080819193657.GE8318@parisc-linux.org> Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-ext4.vger.kernel.org On Tue, Aug 19, 2008 at 01:36:57PM -0600, Matthew Wilcox wrote: > On Tue, Aug 19, 2008 at 03:22:16PM -0400, Josef Bacik wrote: > > On Tue, Aug 19, 2008 at 01:15:08PM -0600, Matthew Wilcox wrote: > > > I don't see any users (in this patch or the next) of people wanting > > > uninterruptible nanosleeps. We shouldn't be introducing new > > > TASK_UNINTERRUPTIBLE users, but instead using TASK_KILLABLE if the user > > > really can't cope with signals in a sensible manner. > > > > Hmm doh, sorry about that the 2/2 patch of this series should be passing 0 not 1 > > since we need to be uninterruptible. > > I think that's taught us that '0' and '1' are insufficiently > descriptive, and we should either be passing in a state (ie > do_nanosleep(x, y, TASK_FOO), or have separate do_nanosleep_killable() > and do_nanosleep_interruptible()). > > > I figured this sort of thing would be used > > by fs's/device drivers where TASK_UNINTERRUPTIBLE is desired. If that is not > > appropriate let me know and I can use TASK_KILLABLE or whatever else the > > preference is. Thanks, > > We should be trying to accommodate the user's wishes wherever possible. > If they say kill -9, they really mean it, and we should stop waiting. > Now, I don't think we should abort the journal_stop(). That's probably > going too far in this instance. But we should stop waiting for other > tasks to join in, and finish up as quickly as possible. Fair enough, I will make the changes in the next round, thanks much. Josef