linux-kernel - Re: [PATCH] dmaengine/ste_dma40: support pm in dma40

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1321507432.1516.176.camel@vkoul-udesk3>
Date:	Thu, 17 Nov 2011 10:53:52 +0530
From:	Vinod Koul <vinod.koul@...el.com>
To:	Narayanan G <narayanan.gopalakrishnan@...ricsson.com>
Cc:	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Linus WALLEIJ <linus.walleij@...ricsson.com>,
	Rabin VINCENT <rabin.vincent@...ricsson.com>
Subject: Re: [PATCH] dmaengine/ste_dma40: support pm in dma40

On Thu, 2011-11-17 at 10:34 +0530, Narayanan G wrote:
> 
> > > +      */
> > > +     if (base->rev == 1)
> > > +             return;
> > > +
> > > +
> > > +     spin_lock_irqsave(&d40c->base->usage_lock, flags);
> > > +
> > > +     d40_power_off(d40c->base, d40c->phy_chan->num);
> > if this does what it says then it is wrong.
> > power_off should be done in your suspend callbacks.
> > same for on as well!!
> 
> Actually, we need to switch off the clocks for the event groups,
> that are not in use. Say, if only evengroup 2 is active, the other
> clocks can be switched off. The clocks for the unused event lines
> need not be on till the runtime suspend is called. May be, I should
> rename this function as d40_power_off_evt_grp(). 
> 
> > > -             d40c->busy = true;
> > > +             if (!d40c->busy) {
> > > +                     d40_usage_inc(d40c);
> > > +                     d40c->busy = true;
> > > +             }
> > well here is problem!
> > You don't need to check busy here, juts call pm_runtime_get(). Power
> on
> > will be take care if it requires resume in your resume callback. No
> need
> > to have checks of busy. You are not properly utilizing functionality
> > provided by runtime_pm
> 
> I have this usage_inc() function mainly for switching on and off the
> clocks for the desired event groups. The busy check here is to ensure
> that we don't need to do the usage_inc() (clock management) in case it
> is
> already on. 
> Is there a way to do this clock management at eventline granularity
> using the pm_runtime() framework? 
runtime pm manages the device power management. It will call your
suspend/resume callbacks when your device is idle/active and manages the
device usage for you.
Within your device you need to do your own management, which should be
linked to channel being active. You should not club the two.

I am not sure if the clock framework can help you on this

-- 
~Vinod

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/