linux-kernel - Re: [PATCH v2 5/5] dmaengine: sh: rz-dmac: Add runtime PM support

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] [day] [month] [year] [list]
Message-ID: <aLmsX1un1tXwILH0@tom-desktop>
Date: Thu, 4 Sep 2025 17:12:48 +0200
From: Tommaso Merciai <tommaso.merciai.xr@...renesas.com>
To: Geert Uytterhoeven <geert@...ux-m68k.org>
Cc: tomm.merciai@...il.com, linux-renesas-soc@...r.kernel.org,
	biju.das.jz@...renesas.com,
	Michael Turquette <mturquette@...libre.com>,
	Stephen Boyd <sboyd@...nel.org>, Vinod Koul <vkoul@...nel.org>,
	Philipp Zabel <p.zabel@...gutronix.de>,
	Wolfram Sang <wsa+renesas@...g-engineering.com>,
	Lad Prabhakar <prabhakar.mahadev-lad.rj@...renesas.com>,
	Fabrizio Castro <fabrizio.castro.jz@...esas.com>,
	Uwe Kleine-König <u.kleine-koenig@...libre.com>,
	linux-clk@...r.kernel.org, linux-kernel@...r.kernel.org,
	dmaengine@...r.kernel.org
Subject: Re: [PATCH v2 5/5] dmaengine: sh: rz-dmac: Add runtime PM support

Hi Geert,

On Thu, Sep 04, 2025 at 05:10:41PM +0200, Geert Uytterhoeven wrote:
> Hi Tommaso,
> 
> On Thu, 4 Sept 2025 at 17:02, Tommaso Merciai
> <tommaso.merciai.xr@...renesas.com> wrote:
> > On Wed, Sep 03, 2025 at 02:17:53PM +0200, Geert Uytterhoeven wrote:
> > > On Wed, 3 Sept 2025 at 10:28, Tommaso Merciai
> > > <tommaso.merciai.xr@...renesas.com> wrote:
> > > > Enable runtime power management in the rz-dmac driver by adding suspend and
> > > > resume callbacks. This ensures the driver can correctly assert and deassert
> > >
> > > This is not really what this patch does: the Runtime PM-related changes
> > > just hide^Wmove reset handling into the runtime callbacks.
> >
> > Agreed.
> >
> > > > the reset control and manage power state transitions during suspend and
> > > > resume. Adding runtime PM support allows the DMA controller to reduce power
> > >
> > > (I assume) This patch does fix resuming from _system_ suspend.
> > >
> > > > consumption when idle and maintain correct operation across system sleep
> > > > states, addressing the previous lack of dynamic power management in the
> > > > driver.
> > >
> > > The driver still does not do dynamic power management: you still call
> > > pm_runtime_resume_and_get() from the driver's probe() .callback, and
> > > call pm_runtime_put() only from the .remove() callback, so the device
> > > is powered all the time.
> > > To implement dynamic power management, you have to change that,
> > > and call pm_runtime_resume_and_get() and pm_runtime_put() from the
> > > .device_alloc_chan_resources() resp. .device_free_chan_resources()
> > > callbacks (see e.g. drivers/dma/sh/rcar-dmac.c).
> >
> > Thanks for the hints!
> > So following your hints we need to:
> >
> >  - call pm_runtime_get_sync() from rz_dmac_alloc_chan_resources()
> >  - call pm_runtime_put() from rz_dmac_free_chan_resources()
> >
> > With that then we can remove pm_runtime_put() from the remove function
> > and add this at the end of the probe function.
> >
> > > > Signed-off-by: Tommaso Merciai <tommaso.merciai.xr@...renesas.com>
> > >
> > > > --- a/drivers/dma/sh/rz-dmac.c
> > > > +++ b/drivers/dma/sh/rz-dmac.c
> > > > @@ -437,6 +437,17 @@ static int rz_dmac_xfer_desc(struct rz_dmac_chan *chan)
> > > >   * DMA engine operations
> > > >   */
> > > >
> > > > +static void rz_dmac_chan_init_all(struct rz_dmac *dmac)
> > > > +{
> > > > +       unsigned int i;
> > > > +
> > > > +       rz_dmac_writel(dmac, DCTRL_DEFAULT, CHANNEL_0_7_COMMON_BASE + DCTRL);
> > > > +       rz_dmac_writel(dmac, DCTRL_DEFAULT, CHANNEL_8_15_COMMON_BASE + DCTRL);
> > > > +
> > > > +       for (i = 0; i < dmac->n_channels; i++)
> > > > +               rz_dmac_ch_writel(&dmac->channels[i], CHCTRL_DEFAULT, CHCTRL, 1);
> > > > +}
> > > > +
> > > >  static int rz_dmac_alloc_chan_resources(struct dma_chan *chan)
> > > >  {
> > > >         struct rz_dmac_chan *channel = to_rz_dmac_chan(chan);
> > > > @@ -970,10 +981,6 @@ static int rz_dmac_probe(struct platform_device *pdev)
> > > >                 goto err_pm_disable;
> > > >         }
> > > >
> > > > -       ret = reset_control_deassert(dmac->rstc);
> > > > -       if (ret)
> > > > -               goto err_pm_runtime_put;
> > > > -
> > > >         for (i = 0; i < dmac->n_channels; i++) {
> > > >                 ret = rz_dmac_chan_probe(dmac, &dmac->channels[i], i);
> > > >                 if (ret < 0)
> > > > @@ -1028,8 +1035,6 @@ static int rz_dmac_probe(struct platform_device *pdev)
> > > >                                   channel->lmdesc.base_dma);
> > > >         }
> > > >
> > > > -       reset_control_assert(dmac->rstc);
> > > > -err_pm_runtime_put:
> > > >         pm_runtime_put(&pdev->dev);
> > > >  err_pm_disable:
> > > >         pm_runtime_disable(&pdev->dev);
> > > > @@ -1052,13 +1057,50 @@ static void rz_dmac_remove(struct platform_device *pdev)
> > > >                                   channel->lmdesc.base,
> > > >                                   channel->lmdesc.base_dma);
> > > >         }
> > > > -       reset_control_assert(dmac->rstc);
> > > >         pm_runtime_put(&pdev->dev);
> > > >         pm_runtime_disable(&pdev->dev);
> > > >
> > > >         platform_device_put(dmac->icu.pdev);
> > > >  }
> > > >
> > > > +static int rz_dmac_runtime_suspend(struct device *dev)
> > > > +{
> > > > +       struct rz_dmac *dmac = dev_get_drvdata(dev);
> > > > +
> > > > +       return reset_control_assert(dmac->rstc);
> > >
> > > Do you really want to reset the device (and thus loose register state)
> > > each and every time the device is runtime-suspended?  For now it doesn't
> > > matter much, but once you implement real dynamic power management,
> > > it does.
> > > I think the reset handling should be moved to the system suspend/resume
> > > callbacks.
> >
> > Agreed. With above changes maybe we can move all into
> > NOIRQ_SYSTEM_SLEEP_PM_OPS(rz_dmac_suspend, rz_dmac_resume)
> > With your suggested changes I'm not sure if pm_runtime_ops are really needed.
> 
> After these changes, you indeed no longer need any pm_runtime_ops.

Thank you for the quick feedback!

Kind Regards,
Tommaso

> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 
> -- 
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@...ux-m68k.org
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds