From mboxrd@z Thu Jan 1 00:00:00 1970 From: Vinod Koul Subject: Re: [PATCH v8 1/5] PM / Runtime: Add getter for querying the IRQ safe option Date: Mon, 3 Nov 2014 21:57:28 +0530 Message-ID: <20141103162728.GB1870@intel.com> References: <1413795888-18559-1-git-send-email-k.kozlowski@samsung.com> <20141031142241.GA17547@amd> <1414766416.6537.6.camel@AMDC1943> <3452823.YQ5FR2FtKz@avalon> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Content-Disposition: inline In-Reply-To: <3452823.YQ5FR2FtKz@avalon> Sender: linux-kernel-owner@vger.kernel.org To: Laurent Pinchart Cc: Krzysztof Kozlowski , Pavel Machek , "Rafael J. Wysocki" , Len Brown , Jonathan Corbet , Russell King , Dan Williams , Ulf Hansson , Alan Stern , linux-pm@vger.kernel.org, linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org, dmaengine@vger.kernel.org, Lars-Peter Clausen , Michal Simek , Kyungmin Park , Marek Szyprowski , Bartlomiej Zolnierkiewicz List-Id: linux-pm@vger.kernel.org On Sat, Nov 01, 2014 at 02:29:42AM +0200, Laurent Pinchart wrote: > Hi Krzysztof, >=20 > On Friday 31 October 2014 15:40:16 Krzysztof Kozlowski wrote: > > On pi=C4=85, 2014-10-31 at 15:22 +0100, Pavel Machek wrote: > > > On Fri 2014-10-31 10:14:55, Krzysztof Kozlowski wrote: > > >> On pon, 2014-10-20 at 11:04 +0200, Krzysztof Kozlowski wrote: > > >>> Add a simple getter pm_runtime_is_irq_safe() for querying wheth= er > > >>> runtime PM IRQ safe was set or not. > > >>>=20 > > >>> Various bus drivers implementing runtime PM may use choose to s= uspend > > >>> differently based on IRQ safeness status of child driver (e.g. = do not > > >>> unprepare the clock if IRQ safe is not set). > > >>>=20 > > >>> Signed-off-by: Krzysztof Kozlowski > > >>> Reviewed-by: Ulf Hansson > > >>=20 > > >> Rafael, Len, Pavel, > > >>=20 > > >> Is proposed API ok? Do you have any comments? > > >>=20 > > >> I'll upload whole patchset to Russell's patch tracking system. H= owever > > >> an ack from PM maintainer is probably needed. > > >=20 > > > I don't like the API. Having callbacks work in different context = (irq > > > / noirq) based on what another function reports is ugly. > > >=20 > > > What is the penalty if we always decide callbacks are not IRQ saf= e? > >=20 > > Then pm_runtime_get_sync() could not be called in atomic context. T= he > > pl330 runtime PM would have to be completely reworked because one > > pm_runtime_get_sync() is called in device_issue_pending which canno= t > > sleep (at least in non preemptible kernels). Probably this can be s= olved > > some way... >=20 > Many other drivers suffer from the same problem. While I won't reject= your=20 > proposed fix, I would prefer a more generic approach. >=20 > One option that has been discussed previously was to use a work queue= to delay=20 > starting the DMA transfer to an interruptible context where=20 > pm_runtime_get_sync() could be called. However, as Russell pointed ou= t [1],=20 > even that won't work in all cases as the DMA slave might need the tra= nsfer to=20 > be started before enabling part of its hardware (OMAP audio seem to b= e such a=20 > case). >=20 > I've heard a rumor of a possible DMA engine rework to forbid calling = the=20 > descriptor preparation API from atomic context. This could be used as= a base=20 > to implement runtime PM, as DMA slave drivers should not prepare desc= riptors=20 > if they don't need to use them. However that's a long term plan, and = we need a=20 > solution sooner than that. Well it is not a rumour :) I have been contemplating that now that async_tx will be killed so we d= ont have to worry about that usage. For the slave dma usage, we can change = the prepare API to be non atomic. I think the users will be okay with appro= ach. This way drivers can use runtime pm calls in prepare. --=20 ~Vinod >=20 > I've been toying with the idea of adding explicit open/close (or what= ever we=20 > would call them) operations to the DMA engine API. Those would be use= d by DMA=20 > slave drivers to signal that they will start/stop using the DMA engin= e. >=20 > If (1) we must start the DMA synchronously with a DMA slave call, (2)= need to=20 > sleep to handle PM, and (3) don't want to keep the DMA engine powered= for as=20 > long as one channel is requested, then we need to turn at least prepa= ration as=20 > not callable in atomic context, or introduce a new operation. >=20 > [1] http://www.spinics.net/lists/dmaengine/msg01548.html >=20 > > >>> --- a/Documentation/power/runtime_pm.txt > > >>> +++ b/Documentation/power/runtime_pm.txt > > >>>=20 > > >>> @@ -468,6 +468,10 @@ drivers/base/power/runtime.c and > > >>> include/linux/pm_runtime.h: > > >>> - set the power.irq_safe flag for the device, causing the > > >>> runtime-PM > > >>> =20 > > >>> callbacks to be invoked with interrupts off > > >>>=20 > > >>> + bool pm_runtime_is_irq_safe(struct device *dev); > > >>> + - return true if power.irq_safe flag was set for the devic= e, > > >>> causing > > >>> + the runtime-PM callbacks to be invoked with interrupts o= ff > > >>> + > > >>>=20 > > >>> void pm_runtime_mark_last_busy(struct device *dev); > > >>> =20 > > >>> - set the power.last_busy field to the current time >=20 > --=20 > Regards, >=20 > Laurent Pinchart >=20 --=20