linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Daniel Kurtz <djkurtz@chromium.org>
To: Mark Yao <mark.yao@rock-chips.com>
Cc: David Airlie <airlied@linux.ie>, Daniel Vetter <daniel@ffwll.ch>,
	Rob Clark <robdclark@gmail.com>,
	Philipp Zabel <p.zabel@pengutronix.de>,
	dri-devel <dri-devel@lists.freedesktop.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"open list:ARM/Rockchip SoC..."
	<linux-rockchip@lists.infradead.org>
Subject: Re: [PATCH] drm/rockchip: vop: power off until vop standby take effect
Date: Mon, 2 Feb 2015 10:07:21 +0800	[thread overview]
Message-ID: <CAGS+omA70JADHR3PhGq-qC8MYobQjsR91MLw6RXdQ6xag0UQWg@mail.gmail.com> (raw)
In-Reply-To: <1422693698-21944-1-git-send-email-mark.yao@rock-chips.com>

Hi Mark, Heiko,

On Sat, Jan 31, 2015 at 4:41 PM, Mark Yao <mark.yao@rock-chips.com> wrote:
> Vop standby will take effect end of current frame,
> if dsp_hold_valid_irq happen, it means vop standby complete.
>
> we must wait standby complete when we want to disable aclk,
> if not, memory bus maybe dead.
>
> Signed-off-by: Mark Yao <mark.yao@rock-chips.com>
> ---
>  drivers/gpu/drm/rockchip/rockchip_drm_vop.c |   76 ++++++++++++++++++++++-----
>  1 file changed, 63 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c
> index fb25836..47ea61f 100644
> --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c
> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c
> @@ -89,6 +89,7 @@ struct vop {
>         /* mutex vsync_ work */
>         struct mutex vsync_mutex;
>         bool vsync_work_pending;
> +       struct completion dsp_hold_completion;
>
>         const struct vop_data *data;
>
> @@ -382,6 +383,34 @@ static bool is_alpha_support(uint32_t format)
>         }
>  }
>
> +static void vop_dsp_hold_valid_irq_enable(struct vop *vop)
> +{
> +       unsigned long flags;
> +
> +       BUG_ON(!vop->is_enabled);

Re: Heiko "use a WARN_ON":

If the VOP clock is off, then the system will just hang when trying to
write the VOP register so in this case, BUG_ON gives a more reliable
crash dump than the hang.

> +
> +       spin_lock_irqsave(&vop->irq_lock, flags);
> +
> +       vop_mask_write(vop, INTR_CTRL0, DSP_HOLD_VALID_INTR_MASK,
> +                      DSP_HOLD_VALID_INTR_EN(1));
> +
> +       spin_unlock_irqrestore(&vop->irq_lock, flags);
> +}
> +
> +static void vop_dsp_hold_valid_irq_disable(struct vop *vop)
> +{
> +       unsigned long flags;
> +
> +       BUG_ON(!vop->is_enabled);
> +
> +       spin_lock_irqsave(&vop->irq_lock, flags);
> +
> +       vop_mask_write(vop, INTR_CTRL0, DSP_HOLD_VALID_INTR_MASK,
> +                      DSP_HOLD_VALID_INTR_EN(0));
> +
> +       spin_unlock_irqrestore(&vop->irq_lock, flags);
> +}
> +
>  static void vop_enable(struct drm_crtc *crtc)
>  {
>         struct vop *vop = to_vop(crtc);
> @@ -454,26 +483,36 @@ static void vop_disable(struct drm_crtc *crtc)
>
>         drm_vblank_off(crtc->dev, vop->pipe);
>
> -       disable_irq(vop->irq);
> -
>         /*
> -        * TODO: Since standby doesn't take effect until the next vblank,
> -        * when we turn off dclk below, the vop is probably still active.
> +        * Vop standby will take effect until end of current frame,

"Vop standby will take effect at end of current frame"

> +        * if dsp hold valid irq happen, it means standby complete.
> +        *
> +        * we must wait standby complete when we want to disable aclk,
> +        * if not, memory bus maybe dead.
>          */
> +       reinit_completion(&vop->dsp_hold_completion);
> +       vop_dsp_hold_valid_irq_enable(vop);
> +
>         spin_lock(&vop->reg_lock);
>
>         VOP_CTRL_SET(vop, standby, 1);
>
>         spin_unlock(&vop->reg_lock);
>
> +       wait_for_completion(&vop->dsp_hold_completion);
> +
> +       vop_dsp_hold_valid_irq_disable(vop);
> +
> +       disable_irq(vop->irq);
> +
>         vop->is_enabled = false;
> +
>         /*
> -        * disable dclk to stop frame scan, so we can safely detach iommu,
> +        * vop standby complete, so iommu detach is safe.
>          */
> -       clk_disable(vop->dclk);
> -
>         rockchip_drm_dma_detach_device(vop->drm_dev, vop->dev);
>
> +       clk_disable(vop->dclk);
>         clk_disable(vop->aclk);
>         clk_disable(vop->hclk);
>  }
> @@ -1086,6 +1125,7 @@ static irqreturn_t vop_isr(int irq, void *data)
>         struct vop *vop = data;
>         uint32_t intr0_reg, active_irqs;
>         unsigned long flags;
> +       int ret = IRQ_NONE;
>
>         /*
>          * INTR_CTRL0 register has interrupt status, enable and clear bits, we
> @@ -1104,15 +1144,24 @@ static irqreturn_t vop_isr(int irq, void *data)
>         if (!active_irqs)
>                 return IRQ_NONE;
>
> -       /* Only Frame Start Interrupt is enabled; other irqs are spurious. */
> -       if (!(active_irqs & FS_INTR)) {
> -               DRM_ERROR("Unknown VOP IRQs: %#02x\n", active_irqs);
> -               return IRQ_NONE;
> +       if (active_irqs & DSP_HOLD_VALID_INTR) {
> +               if (!completion_done(&vop->dsp_hold_completion))

Why is this "completion_done" check needed?
I guess it doesn't help, but isn't strictly needed, since the
dsp_hold_completion is a one shot, right?

In any case, this should work as it is, so:

Reviewed-by: Daniel Kurtz <djkurtz@chromium.org>

> +                       complete(&vop->dsp_hold_completion);
> +               active_irqs &= ~DSP_HOLD_VALID_INTR;
> +               ret = IRQ_HANDLED;
>         }
>
> -       drm_handle_vblank(vop->drm_dev, vop->pipe);
> +       if (active_irqs & FS_INTR) {
> +               drm_handle_vblank(vop->drm_dev, vop->pipe);
> +               active_irqs &= ~FS_INTR;
> +               ret = (vop->vsync_work_pending) ? IRQ_WAKE_THREAD : IRQ_HANDLED;
> +       }
>
> -       return (vop->vsync_work_pending) ? IRQ_WAKE_THREAD : IRQ_HANDLED;
> +       /* Unhandled irqs are spurious. */
> +       if (active_irqs)
> +               DRM_ERROR("Unknown VOP IRQs: %#02x\n", active_irqs);
> +
> +       return ret;
>  }
>
>  static int vop_create_crtc(struct vop *vop)
> @@ -1194,6 +1243,7 @@ static int vop_create_crtc(struct vop *vop)
>                 goto err_cleanup_crtc;
>         }
>
> +       init_completion(&vop->dsp_hold_completion);
>         crtc->port = port;
>         vop->pipe = drm_crtc_index(crtc);
>         rockchip_register_crtc_funcs(drm_dev, &private_crtc_funcs, vop->pipe);
> --
> 1.7.9.5
>
>

  parent reply	other threads:[~2015-02-02  2:07 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-01-31  8:41 [PATCH] drm/rockchip: vop: power off until vop standby take effect Mark Yao
2015-01-31 12:49 ` Heiko Stübner
2015-02-02  0:45   ` Mark yao
2015-02-02  2:07 ` Daniel Kurtz [this message]
2015-02-02  2:30   ` Mark yao
2015-02-02  7:53     ` Daniel Vetter
2015-02-04  3:38       ` Mark yao
2015-02-04  3:48         ` Daniel Kurtz
2015-02-04  3:56           ` Mark yao

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAGS+omA70JADHR3PhGq-qC8MYobQjsR91MLw6RXdQ6xag0UQWg@mail.gmail.com \
    --to=djkurtz@chromium.org \
    --cc=airlied@linux.ie \
    --cc=daniel@ffwll.ch \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-rockchip@lists.infradead.org \
    --cc=mark.yao@rock-chips.com \
    --cc=p.zabel@pengutronix.de \
    --cc=robdclark@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).