From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757212Ab2ASCgD (ORCPT ); Wed, 18 Jan 2012 21:36:03 -0500 Received: from mail-tul01m020-f174.google.com ([209.85.214.174]:38915 "EHLO mail-tul01m020-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755574Ab2ASCf5 convert rfc822-to-8bit (ORCPT ); Wed, 18 Jan 2012 21:35:57 -0500 MIME-Version: 1.0 In-Reply-To: <20120119013731.GK2431@linux.vnet.ibm.com> References: <1326826563-32215-1-git-send-email-sjg@chromium.org> <20120118224304.GJ2431@linux.vnet.ibm.com> <201201190102.58788.rjw@sisk.pl> <20120119013731.GK2431@linux.vnet.ibm.com> Date: Wed, 18 Jan 2012 18:35:56 -0800 X-Google-Sender-Auth: g_ZmqtRdIUB67kdiGy4jTo8_5E4 Message-ID: Subject: Re: [PATCH 3/3] serial: 8250: Add a wakeup_capable module param From: Simon Glass To: paulmck@linux.vnet.ibm.com Cc: "Rafael J. Wysocki" , Alan Cox , LKML , Greg Kroah-Hartman , linux-serial@vger.kernel.org X-System-Of-Record: true Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Paul, On Wed, Jan 18, 2012 at 5:37 PM, Paul E. McKenney wrote: > On Thu, Jan 19, 2012 at 01:02:58AM +0100, Rafael J. Wysocki wrote: >> On Wednesday, January 18, 2012, Paul E. McKenney wrote: >> > On Wed, Jan 18, 2012 at 02:15:59PM -0800, Simon Glass wrote: >> > > Hi Paul, >> > > >> > > On Wed, Jan 18, 2012 at 1:42 PM, Paul E. McKenney >> > > wrote: >> > > > On Wed, Jan 18, 2012 at 01:08:13PM -0800, Simon Glass wrote: >> > > >> [+cc Rafael J. Wysocki who I think wrote the wakeup.c code] >> > > >> >> > > >> Hi Alan, Paul, >> > > >> >> > > >> On Tue, Jan 17, 2012 at 8:17 PM, Paul E. McKenney >> > > >> wrote: >> > > >> > On Tue, Jan 17, 2012 at 08:10:36PM +0000, Alan Cox wrote: >> > > >> >> On Tue, 17 Jan 2012 10:56:03 -0800 >> > > >> >> Simon Glass wrote: >> > > >> >> >> > > >> >> > Since serial_core now does not make serial ports wake-up capable by >> > > >> >> > default, add a parameter to support this feature in the 8250 UART. >> > > >> >> > This is the only UART where I think this feature is useful. >> > > >> >> >> > > >> >> NAK >> > > >> >> >> > > >> >> Things should just work for users. Magic parameters is not an >> > > >> >> improvement. If its a performance problem someone needs to fix the rcu >> > > >> >> sync overhead or stop using rcu on that path. >> > > >> >> > > >> OK fair enough, I agree. Every level I move down the source tree >> > > >> affects more people though. >> > > >> >> > > >> > >> > > >> > I must say that I lack context here, even after looking at the patch, >> > > >> > but the synchronize_rcu_expedited() primitives can be used if the latency >> > > >> > of synchronize_rcu() is too large. >> > > >> > >> > > >> >> > > >> Let me provide a bit of context. The serial_core code seems to be the >> > > >> only place in the kernel that does this: >> > > >> >> > > >>               device_init_wakeup(tty_dev, 1); >> > > >>               device_set_wakeup_enable(tty_dev, 0); >> > > >> >> > > >> The first call makes the device wakeup capable and enables wakeup, The >> > > >> second call disabled wakeup. >> > > >> >> > > >> The code that removes the wakeup source looks like this: >> > > >> >> > > >> void wakeup_source_remove(struct wakeup_source *ws) >> > > >> { >> > > >>       if (WARN_ON(!ws)) >> > > >>               return; >> > > >> >> > > >>       spin_lock_irq(&events_lock); >> > > >>       list_del_rcu(&ws->entry); >> > > >>       spin_unlock_irq(&events_lock); >> > > >>       synchronize_rcu(); >> > > >> } >> > > >> >> > > >> The sync is there because we are about to destroy the actual ws >> > > >> structure (in wakeup_source_destroy()). I wonder if it should be in >> > > >> wakeup_source_destroy() but that wouldn't help me anyway. >> > > >> >> > > >> synchronize_rcu_expedited() is a bit faster but not really fast >> > > >> enough. Anyway surely people will complain if I put this in the wakeup >> > > >> code - it will affect all wakeup users. It seems to me that the right >> > > >> solution is to avoid enabling and then immediately disabling wakeup. >> > > > >> > > > Hmmm...  What hardware are you running this one?  Normally, >> > > > synchronize_rcu_expedited() will be a couple of orders of magnitude >> > > > faster than synchronize_rcu(). >> > > > >> > > >> I assume we can't and shouldn't change device_init_wakeup() . We could >> > > >> add a call like device_init_wakeup_disabled() which makes the device >> > > >> wakeup capable but does not actually enable it. Does that work? >> > > > >> > > > If the only reason for the synchronize_rcu() is to defer the pair of >> > > > kfree()s in wakeup_source_destroy(), then another possible approach >> > > > would be to remove the synchronize_rcu() from wakeup_source_remove() >> > > > and then use call_rcu() to defer the two kfree()s. >> > > > >> > > > If this is a reasonable change to make, the approach is as follows: >> > > > >> > > > 1.      Add a struct rcu_head to wakeup_source, call it "rcu". >> > > >        Or adjust the following to suit your choice of name. >> > > > >> > > > 2.      Replace the pair of kfree()s with: >> > > > >> > > >                call_rcu(&ws->rcu, wakeup_source_destroy_rcu); >> > > > >> > > > 3.      Create the wakeup_source_destroy_rcu() as follows: >> > > > >> > > >        static void wakeup_source_destroy_rcu(struct rcu_head *head) >> > > >        { >> > > >                struct wakeup_source *ws = >> > > >                        container_of(head, struct wakeup_source, rcu); >> > > > >> > > >                kfree(ws->name); >> > > >                kfree(ws); >> > > >        } >> > > > >> > > > Of course, this assumes that it is OK for wakeup_source_unregister() >> > > > to return before the memory is freed up.  This often is OK, but there >> > > > are some cases where the caller requires that there be no further >> > > > RCU readers with access to the old data.  In these cases, you really >> > > > do need the wait. >> > > >> > > Thanks very much for that. I'm not sure if it is a reasonable change, >> > > but it does bug me that we add it to a data structure knowing that we >> > > will immediately remove it! >> > > >> > > >From what I can see, making a device wakeup-enabled mostly happens on >> > > init or in response to a request to the driver (presumably from user >> > > space). In the latter case I suspect the synchronise_rcu() is fine. In >> > > the former it feels like we should make up our minds which of the >> > > three options is required (incapable, capable but not enabled, capable >> > > and enabled). >> > > >> > > I will try a patch first based on splitting the two options (capable >> > > and enable) and see if that get a NAK. >> > > >> > > Then I will come back to your solution - it seems fine to me and not a >> > > lot of code. Do we have to worry about someone enabling, disabled, >> > > enabling and then disabling wakeup quickly? Will this method break in >> > > that case if the second call to call_rcu() uses the same wc->rcu? >> > >> > There are a couple of questions here, let me take them one at a time: >> > >> > 1.  If you just disabled, can you immediately re-enable? >> > >> >     The answer is "yes".  The reason that this works is that you >> >     allocate a new structure for the re-enabling, and that new >> >     structure has its own rcu_head field. >> > >> > 2.  If you repeatedly disable and re-enable in a tight loop, >> >     can this cause problems? >> > >> >     The answer to this is also "yes" -- you can run the system >> >     out of memory doing that.  However, there are a number of >> >     simple ways to avoid this problem: >> > >> >     a.      Do a synchronize_rcu() on every (say) thousandth >> >             disable operation. >> > >> >     b.      As above, but only do the synchronize_rcu() if >> >             all 1,000 disable operations occurred within >> >             (say) a second of each other. >> > >> >     c.      As above, but actually count the number of >> >             pending call_rcu() callbacks. >> > >> >     Both (a) and (b) can be carried out on a per-CPU basis if there >> >     is no convenient locked structure in which to track the state. >> >     You cannot carry (c) out on a per-CPU basis because RCU callbacks >> >     can sometimes be invoked on a different CPU from the one that >> >     call_rcu()ed them.  Rare, but it can happen. >> > >> >     I would expect that option (a) would work in almost all cases. >> > >> > If this can be exercised freely from user space, then you probably >> > really do need #2 above. >> >> Yes, you can, but then I'd say it's not necessary for user space to >> be able to carry that out in a tight loop.  So, it seems, alternatively, >> we could make that loop a bit less tight, e.g. by adding an arbitrary >> sleep to the user space interface for the "disable" case. > > Good point, that would work just as well and be simpler. OK, well I am expecting that this will now be a very small patch to change just serial_core. Thanks for your help with this. Regards, Simon