On Wed, Oct 07, 2015 at 12:05:37PM +0800, Yakir Yang wrote:
>
>
> On 08/09/2015 12:04 AM, Russell King wrote:
> >The dw_hdmi enable/disable handling is particularly weak in several
> >regards:
> >* The hotplug interrupt could call hdmi_poweron() or hdmi_poweroff()
> > while DRM is setting a mode, which could race with a mode being set.
> >* Hotplug will always re-enable the phy whenever it detects an active
> > hotplug signal, even if DRM has disabled the output.
> >
> >Resolve all of these by introducing a mutex to prevent races, and a
> >state-tracking bool so we know whether DRM wishes the output to be
> >enabled. We choose to use our own mutex rather than ->struct_mutex
> >so that we can still process interrupts in a timely fashion.
> >
> >Signed-off-by: Russell King <rmk+kernel at arm.linux.org.uk>
> >---
> > drivers/gpu/drm/bridge/dw_hdmi.c | 29 ++++++++++++++++++++++-------
> > 1 file changed, 22 insertions(+), 7 deletions(-)
> >
> >diff --git a/drivers/gpu/drm/bridge/dw_hdmi.c
> >b/drivers/gpu/drm/bridge/dw_hdmi.c
> >index 7b8a4e942a71..0ee188930d26 100644
> >--- a/drivers/gpu/drm/bridge/dw_hdmi.c
> >+++ b/drivers/gpu/drm/bridge/dw_hdmi.c
> >@@ -125,6 +125,9 @@ struct dw_hdmi {
> > bool sink_is_hdmi;
> > bool sink_has_audio;
> >+ struct mutex mutex; /* for state below and previous_mode */
> >+ bool disabled; /* DRM has disabled our bridge */
> >+
> > spinlock_t audio_lock;
> > struct mutex audio_mutex;
> > unsigned int sample_rate;
> >@@ -1389,8 +1392,12 @@ static void dw_hdmi_bridge_mode_set(struct drm_bridge
> >*bridge,
> > {
> > struct dw_hdmi *hdmi = bridge->driver_private;
> >+ mutex_lock(&hdmi->mutex);
> >+
> > /* Store the display mode for plugin/DKMS poweron events */
> > memcpy(&hdmi->previous_mode, mode, sizeof(hdmi->previous_mode));
> >+
> >+ mutex_unlock(&hdmi->mutex);
> > }
> > static bool dw_hdmi_bridge_mode_fixup(struct drm_bridge *bridge,
> >@@ -1404,14 +1411,20 @@ static void dw_hdmi_bridge_disable(struct drm_bridge
> >*bridge)
> > {
> > struct dw_hdmi *hdmi = bridge->driver_private;
> >+ mutex_lock(&hdmi->mutex);
> >+ hdmi->disabled = true;
> > dw_hdmi_poweroff(hdmi);
> >+ mutex_unlock(&hdmi->mutex);
> > }
> > static void dw_hdmi_bridge_enable(struct drm_bridge *bridge)
> > {
> > struct dw_hdmi *hdmi = bridge->driver_private;
> >+ mutex_lock(&hdmi->mutex);
> > dw_hdmi_poweron(hdmi);
> >+ hdmi->disabled = false;
> >+ mutex_unlock(&hdmi->mutex);
> > }
> > static void dw_hdmi_bridge_nop(struct drm_bridge *bridge)
> >@@ -1534,20 +1547,20 @@ static irqreturn_t dw_hdmi_irq(int irq, void *dev_id)
> > phy_int_pol = hdmi_readb(hdmi, HDMI_PHY_POL0);
> > if (intr_stat & HDMI_IH_PHY_STAT0_HPD) {
> >+ hdmi_modb(hdmi, ~phy_int_pol, HDMI_PHY_HPD, HDMI_PHY_POL0);
> >+ mutex_lock(&hdmi->mutex);
> > if (phy_int_pol & HDMI_PHY_HPD) {
> > dev_dbg(hdmi->dev, "EVENT=plugin\n");
> >- hdmi_modb(hdmi, 0, HDMI_PHY_HPD, HDMI_PHY_POL0);
> >-
> >- dw_hdmi_poweron(hdmi);
> >+ if (!hdmi->disabled)
> >+ dw_hdmi_poweron(hdmi);
> > } else {
> > dev_dbg(hdmi->dev, "EVENT=plugout\n");
> >- hdmi_modb(hdmi, HDMI_PHY_HPD, HDMI_PHY_HPD,
> >- HDMI_PHY_POL0);
> >-
> >- dw_hdmi_poweroff(hdmi);
> >+ if (!hdmi->disabled)
> >+ dw_hdmi_poweroff(hdmi);
>
> Just like my reply on 08/12, I thought this could be removed, so
> poweron/poweroff would only be called with bridge->enable/
> bridge->disable, them maybe no need mutex here.
The bridge enable/disable methods do not get called on hotplug changes.
[ 1.363011] dwhdmi-imx 120000.hdmi: Detected HDMI controller
0x13:0xa:0xa0:0xc1
[ 1.371341] dwhdmi-imx 120000.hdmi: dw_hdmi_irq(I:RX----,HPD P:RX3210,HPD
S:RX----,HPD)
[ 1.381345] imx-drm display-subsystem: bound 120000.hdmi (ops
dw_hdmi_imx_ops)
[ 1.448691] dwhdmi-imx 120000.hdmi: dw_hdmi_bridge_disable()
[ 1.450963] dwhdmi-imx 120000.hdmi: dw_hdmi_bridge_enable()
and then unplugging and re-plugging the HDMI cable:
[ 68.307505] dwhdmi-imx 120000.hdmi: dw_hdmi_irq(I:RX----,HPD P:RX3210,---
S:RX----,---)
[ 73.813970] dwhdmi-imx 120000.hdmi: dw_hdmi_irq(I:RX----,HPD P:RX3210,HPD
S:RX----,HPD)
As you can see, during the period of disconnection for five seconds,
dw_hdmi_bridge_disable() was not called.
So, without the code above, we'd be needlessly wasting power with the
bridge enabled, trying to drive a disconnected display.
--
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.
