diff mbox series

[v4,24/80] drm/omap: dsi: move TE GPIO handling into core

Message ID 20201124124538.660710-25-tomi.valkeinen@ti.com (mailing list archive)
State New
Headers show
Series Convert DSI code to use drm_mipi_dsi and drm_panel | expand

Commit Message

Tomi Valkeinen Nov. 24, 2020, 12:44 p.m. UTC
From: Sebastian Reichel <sebastian.reichel@collabora.com>

In preparation for removing custom DSS calls from the DSI
panel driver, this moves support for external tearing event
GPIOs into the DSI host driver. This way tearing events are
always handled in the core resulting in simplification of
the panel drivers.

The TE GPIO acquisition follows works in the same way as the
exynos DSI implementation.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
---
 .../gpu/drm/omapdrm/displays/panel-dsi-cm.c   | 103 +-----------
 drivers/gpu/drm/omapdrm/dss/dsi.c             | 146 ++++++++++++++++--
 2 files changed, 141 insertions(+), 108 deletions(-)

Comments

Tony Lindgren Feb. 8, 2021, 5:55 p.m. UTC | #1
Hi,

* Tomi Valkeinen <tomi.valkeinen@ti.com> [201124 12:47]:
> From: Sebastian Reichel <sebastian.reichel@collabora.com>
> 
> In preparation for removing custom DSS calls from the DSI
> panel driver, this moves support for external tearing event
> GPIOs into the DSI host driver. This way tearing events are
> always handled in the core resulting in simplification of
> the panel drivers.
> 
> The TE GPIO acquisition follows works in the same way as the
> exynos DSI implementation.

Looks like this patch causes the following warnings:

DSI: omapdss DSI error: Failed to receive BTA
DSI: omapdss DSI error: bta sync failed
DSI: omapdss DSI error: vc(0) busy when trying to config for VP
DSI: omapdss DSI error: Failed to receive BTA
DSI: omapdss DSI error: bta sync failed
DSI: omapdss DSI error: vc(0) busy when trying to config for VP
DSI: omapdss DSI error: Failed to receive BTA
DSI: omapdss DSI error: bta sync failed
DSI: omapdss DSI error: vc(0) busy when trying to config for VP
...

Any ideas? The display works for me despite the constant
warnings.

Regards,

Tony
Tomi Valkeinen Feb. 11, 2021, 7:34 a.m. UTC | #2
On 08/02/2021 19:55, Tony Lindgren wrote:
> Hi,
> 
> * Tomi Valkeinen <tomi.valkeinen@ti.com> [201124 12:47]:
>> From: Sebastian Reichel <sebastian.reichel@collabora.com>
>>
>> In preparation for removing custom DSS calls from the DSI
>> panel driver, this moves support for external tearing event
>> GPIOs into the DSI host driver. This way tearing events are
>> always handled in the core resulting in simplification of
>> the panel drivers.
>>
>> The TE GPIO acquisition follows works in the same way as the
>> exynos DSI implementation.
> 
> Looks like this patch causes the following warnings:
> 
> DSI: omapdss DSI error: Failed to receive BTA
> DSI: omapdss DSI error: bta sync failed
> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> DSI: omapdss DSI error: Failed to receive BTA
> DSI: omapdss DSI error: bta sync failed
> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> DSI: omapdss DSI error: Failed to receive BTA
> DSI: omapdss DSI error: bta sync failed
> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> ...
> 
> Any ideas? The display works for me despite the constant
> warnings.

Which board is that? Do the errors start right from the boot, or only
after running something in userspace?

 Tomi
Tony Lindgren Feb. 11, 2021, 5:35 p.m. UTC | #3
* Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210211 07:35]:
> On 08/02/2021 19:55, Tony Lindgren wrote:
> > Hi,
> > 
> > * Tomi Valkeinen <tomi.valkeinen@ti.com> [201124 12:47]:
> >> From: Sebastian Reichel <sebastian.reichel@collabora.com>
> >>
> >> In preparation for removing custom DSS calls from the DSI
> >> panel driver, this moves support for external tearing event
> >> GPIOs into the DSI host driver. This way tearing events are
> >> always handled in the core resulting in simplification of
> >> the panel drivers.
> >>
> >> The TE GPIO acquisition follows works in the same way as the
> >> exynos DSI implementation.
> > 
> > Looks like this patch causes the following warnings:
> > 
> > DSI: omapdss DSI error: Failed to receive BTA
> > DSI: omapdss DSI error: bta sync failed
> > DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> > DSI: omapdss DSI error: Failed to receive BTA
> > DSI: omapdss DSI error: bta sync failed
> > DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> > DSI: omapdss DSI error: Failed to receive BTA
> > DSI: omapdss DSI error: bta sync failed
> > DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> > ...
> > 
> > Any ideas? The display works for me despite the constant
> > warnings.
> 
> Which board is that? Do the errors start right from the boot, or only
> after running something in userspace?

This is with droid4, that's about the only device I use with a display
on regular basis. I'm pretty sure some earlier version of Sebastian's
patches worked fine.

Regards,

Tony
Tomi Valkeinen Feb. 17, 2021, 7:42 a.m. UTC | #4
On 11/02/2021 19:35, Tony Lindgren wrote:
> * Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210211 07:35]:
>> On 08/02/2021 19:55, Tony Lindgren wrote:
>>> Hi,
>>>
>>> * Tomi Valkeinen <tomi.valkeinen@ti.com> [201124 12:47]:
>>>> From: Sebastian Reichel <sebastian.reichel@collabora.com>
>>>>
>>>> In preparation for removing custom DSS calls from the DSI
>>>> panel driver, this moves support for external tearing event
>>>> GPIOs into the DSI host driver. This way tearing events are
>>>> always handled in the core resulting in simplification of
>>>> the panel drivers.
>>>>
>>>> The TE GPIO acquisition follows works in the same way as the
>>>> exynos DSI implementation.
>>>
>>> Looks like this patch causes the following warnings:
>>>
>>> DSI: omapdss DSI error: Failed to receive BTA
>>> DSI: omapdss DSI error: bta sync failed
>>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
>>> DSI: omapdss DSI error: Failed to receive BTA
>>> DSI: omapdss DSI error: bta sync failed
>>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
>>> DSI: omapdss DSI error: Failed to receive BTA
>>> DSI: omapdss DSI error: bta sync failed
>>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
>>> ...
>>>
>>> Any ideas? The display works for me despite the constant
>>> warnings.
>>
>> Which board is that? Do the errors start right from the boot, or only
>> after running something in userspace?
> 
> This is with droid4, that's about the only device I use with a display
> on regular basis. I'm pretty sure some earlier version of Sebastian's
> patches worked fine.

OMAP4 SDP doesn't produce these errors and the HW looks rather
identical. Although I noticed something odd there, running kmstest
--flip on the first display works fine, but running on the second
display gets a bit erratic fps. Which is a bit odd as everything should
be identical.

So these errors start from the boot? Or only when running something
specific?

Is there a bootloader that initializes the display?

 Tomi
Tony Lindgren Feb. 18, 2021, 5:57 a.m. UTC | #5
* Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210217 07:42]:
> On 11/02/2021 19:35, Tony Lindgren wrote:
> > * Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210211 07:35]:
> >> On 08/02/2021 19:55, Tony Lindgren wrote:
> >>> Hi,
> >>>
> >>> * Tomi Valkeinen <tomi.valkeinen@ti.com> [201124 12:47]:
> >>>> From: Sebastian Reichel <sebastian.reichel@collabora.com>
> >>>>
> >>>> In preparation for removing custom DSS calls from the DSI
> >>>> panel driver, this moves support for external tearing event
> >>>> GPIOs into the DSI host driver. This way tearing events are
> >>>> always handled in the core resulting in simplification of
> >>>> the panel drivers.
> >>>>
> >>>> The TE GPIO acquisition follows works in the same way as the
> >>>> exynos DSI implementation.
> >>>
> >>> Looks like this patch causes the following warnings:
> >>>
> >>> DSI: omapdss DSI error: Failed to receive BTA
> >>> DSI: omapdss DSI error: bta sync failed
> >>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> >>> DSI: omapdss DSI error: Failed to receive BTA
> >>> DSI: omapdss DSI error: bta sync failed
> >>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> >>> DSI: omapdss DSI error: Failed to receive BTA
> >>> DSI: omapdss DSI error: bta sync failed
> >>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
> >>> ...
> >>>
> >>> Any ideas? The display works for me despite the constant
> >>> warnings.
> >>
> >> Which board is that? Do the errors start right from the boot, or only
> >> after running something in userspace?
> > 
> > This is with droid4, that's about the only device I use with a display
> > on regular basis. I'm pretty sure some earlier version of Sebastian's
> > patches worked fine.
> 
> OMAP4 SDP doesn't produce these errors and the HW looks rather
> identical. Although I noticed something odd there, running kmstest
> --flip on the first display works fine, but running on the second
> display gets a bit erratic fps. Which is a bit odd as everything should
> be identical.

Oh cool that you have those running again/still :) In this case there
is no te-gpios if that might make a difference.

> So these errors start from the boot? Or only when running something
> specific?

They start from the boot when modules are loaded.

> Is there a bootloader that initializes the display?

Yes it boots with kexec.

Regards,

Tony
Tomi Valkeinen Feb. 22, 2021, 8:47 a.m. UTC | #6
On 18/02/2021 07:57, Tony Lindgren wrote:
> * Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210217 07:42]:
>> On 11/02/2021 19:35, Tony Lindgren wrote:
>>> * Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210211 07:35]:
>>>> On 08/02/2021 19:55, Tony Lindgren wrote:
>>>>> Hi,
>>>>>
>>>>> * Tomi Valkeinen <tomi.valkeinen@ti.com> [201124 12:47]:
>>>>>> From: Sebastian Reichel <sebastian.reichel@collabora.com>
>>>>>>
>>>>>> In preparation for removing custom DSS calls from the DSI
>>>>>> panel driver, this moves support for external tearing event
>>>>>> GPIOs into the DSI host driver. This way tearing events are
>>>>>> always handled in the core resulting in simplification of
>>>>>> the panel drivers.
>>>>>>
>>>>>> The TE GPIO acquisition follows works in the same way as the
>>>>>> exynos DSI implementation.
>>>>>
>>>>> Looks like this patch causes the following warnings:
>>>>>
>>>>> DSI: omapdss DSI error: Failed to receive BTA
>>>>> DSI: omapdss DSI error: bta sync failed
>>>>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
>>>>> DSI: omapdss DSI error: Failed to receive BTA
>>>>> DSI: omapdss DSI error: bta sync failed
>>>>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
>>>>> DSI: omapdss DSI error: Failed to receive BTA
>>>>> DSI: omapdss DSI error: bta sync failed
>>>>> DSI: omapdss DSI error: vc(0) busy when trying to config for VP
>>>>> ...
>>>>>
>>>>> Any ideas? The display works for me despite the constant
>>>>> warnings.
>>>>
>>>> Which board is that? Do the errors start right from the boot, or only
>>>> after running something in userspace?
>>>
>>> This is with droid4, that's about the only device I use with a display
>>> on regular basis. I'm pretty sure some earlier version of Sebastian's
>>> patches worked fine.
>>
>> OMAP4 SDP doesn't produce these errors and the HW looks rather
>> identical. Although I noticed something odd there, running kmstest
>> --flip on the first display works fine, but running on the second
>> display gets a bit erratic fps. Which is a bit odd as everything should
>> be identical.
> 
> Oh cool that you have those running again/still :) In this case there
> is no te-gpios if that might make a difference.

No, GPIO TE is not used on OMAP4 SDP either.

>> So these errors start from the boot? Or only when running something
>> specific?
> 
> They start from the boot when modules are loaded.

Normally there are no updates happening unless an userspace app is
running, but I guess you have fb console enabled, with the blinking
cursor which makes the updates.

I usually don't have fbcon enabled, but OMAP4 SDP works fine for me with
fbcon too...

>> Is there a bootloader that initializes the display?
> 
> Yes it boots with kexec.

Is that open source? Can you disable the display setup from the
bootloader? Maybe the DSS or the panel is left into a state that for
whatever reason makes the kernel drivers break.

Or maybe a DSS or DSI reset via SYSCONFIG at probe would help, or panel
reset if it has such a feature.

 Tomi
Tony Lindgren Feb. 25, 2021, 12:46 p.m. UTC | #7
* Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210222 08:47]:
> On 18/02/2021 07:57, Tony Lindgren wrote:
> > Oh cool that you have those running again/still :) In this case there
> > is no te-gpios if that might make a difference.
> 
> No, GPIO TE is not used on OMAP4 SDP either.

OK

> >> So these errors start from the boot? Or only when running something
> >> specific?
> > 
> > They start from the boot when modules are loaded.
> 
> Normally there are no updates happening unless an userspace app is
> running, but I guess you have fb console enabled, with the blinking
> cursor which makes the updates.
> 
> I usually don't have fbcon enabled, but OMAP4 SDP works fine for me with
> fbcon too...

I'm using loadable modules with omap2plus_defconfig in case that makes
a difference. Maybe there's some state preserved somewhere if deferred
probe happens?

> >> Is there a bootloader that initializes the display?
> > 
> > Yes it boots with kexec.
> 
> Is that open source? Can you disable the display setup from the
> bootloader? Maybe the DSS or the panel is left into a state that for
> whatever reason makes the kernel drivers break.

Well that's a signed kernel booting kexec. But it has been working
just fine for years now so I'd rather not blame that.

> Or maybe a DSS or DSI reset via SYSCONFIG at probe would help, or panel
> reset if it has such a feature.

Maybe. But why would the $subject patch cause the errors?

Regards,

Tony
Tomi Valkeinen Feb. 26, 2021, 9:04 a.m. UTC | #8
On 25/02/2021 14:46, Tony Lindgren wrote:
> * Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210222 08:47]:
>> On 18/02/2021 07:57, Tony Lindgren wrote:
>>> Oh cool that you have those running again/still :) In this case there
>>> is no te-gpios if that might make a difference.
>>
>> No, GPIO TE is not used on OMAP4 SDP either.
> 
> OK
> 
>>>> So these errors start from the boot? Or only when running something
>>>> specific?
>>>
>>> They start from the boot when modules are loaded.
>>
>> Normally there are no updates happening unless an userspace app is
>> running, but I guess you have fb console enabled, with the blinking
>> cursor which makes the updates.
>>
>> I usually don't have fbcon enabled, but OMAP4 SDP works fine for me with
>> fbcon too...
> 
> I'm using loadable modules with omap2plus_defconfig in case that makes
> a difference. Maybe there's some state preserved somewhere if deferred
> probe happens?

I don't think the drivers do much unless probe succeeds. I'm also using 
modules.

>>>> Is there a bootloader that initializes the display?
>>>
>>> Yes it boots with kexec.
>>
>> Is that open source? Can you disable the display setup from the
>> bootloader? Maybe the DSS or the panel is left into a state that for
>> whatever reason makes the kernel drivers break.
> 
> Well that's a signed kernel booting kexec. But it has been working
> just fine for years now so I'd rather not blame that.

I'm just looking for the difference with droid4 and omap4 sdp so that I 
could reproduce.

>> Or maybe a DSS or DSI reset via SYSCONFIG at probe would help, or panel
>> reset if it has such a feature.
> 
> Maybe. But why would the $subject patch cause the errors?

Hmm, if I read the code right, TE was not enabled at all before this 
patch. And this patch enables it. So maybe TE has never worked with that 
panel?

You could try changing the enable_te calls to pass false.

Or with the upstream driver, comment out

mipi_dsi_dcs_set_tear_on(ddata->dsi, MIPI_DSI_DCS_TEAR_MODE_VBLANK);

  Tomi
Tony Lindgren Feb. 27, 2021, 7:18 a.m. UTC | #9
* Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> [210226 09:04]:
> Hmm, if I read the code right, TE was not enabled at all before this patch.
> And this patch enables it. So maybe TE has never worked with that panel?
> 
> You could try changing the enable_te calls to pass false.
> 
> Or with the upstream driver, comment out
> 
> mipi_dsi_dcs_set_tear_on(ddata->dsi, MIPI_DSI_DCS_TEAR_MODE_VBLANK);

Yup you're right that's it.

Commenting it out in drivers/gpu/drm/panel/panel-dsi-cm.c makes the
warnings go away.

There are some BTA related comments in the old v3.0.8 based Android
kernel for droid4 panel at [0], but I don't really follow what they
mean. Maybe it's some ordering issue?

Regards,

Tony


[0] https://github.com/NotKit/android_kernel_motorola_omap4-common/blob/hybris-11.0/drivers/video/omap2/displays/panel-mapphone.c
diff mbox series

Patch

diff --git a/drivers/gpu/drm/omapdrm/displays/panel-dsi-cm.c b/drivers/gpu/drm/omapdrm/displays/panel-dsi-cm.c
index aef5d59181fa..35810ec65b7a 100644
--- a/drivers/gpu/drm/omapdrm/displays/panel-dsi-cm.c
+++ b/drivers/gpu/drm/omapdrm/displays/panel-dsi-cm.c
@@ -54,7 +54,6 @@  struct panel_drv_data {
 
 	/* panel HW configuration from DT or platform data */
 	struct gpio_desc *reset_gpio;
-	struct gpio_desc *ext_te_gpio;
 
 	struct regulator_bulk_data supplies[DCS_REGULATOR_SUPPLY_NUM];
 
@@ -66,12 +65,6 @@  struct panel_drv_data {
 	/* runtime variables */
 	bool enabled;
 
-	bool te_enabled;
-
-	atomic_t do_update;
-
-	struct delayed_work te_timeout_work;
-
 	bool intro_printed;
 
 	struct workqueue_struct *workqueue;
@@ -83,8 +76,6 @@  struct panel_drv_data {
 
 #define to_panel_data(p) container_of(p, struct panel_drv_data, dssdev)
 
-static irqreturn_t dsicm_te_isr(int irq, void *data);
-static void dsicm_te_timeout_work_callback(struct work_struct *work);
 static int _dsicm_enable_te(struct panel_drv_data *ddata, bool enable);
 
 static int dsicm_panel_reset(struct panel_drv_data *ddata);
@@ -240,9 +231,6 @@  static int dsicm_enter_ulps(struct panel_drv_data *ddata)
 	if (r)
 		goto err;
 
-	if (ddata->ext_te_gpio)
-		disable_irq(gpiod_to_irq(ddata->ext_te_gpio));
-
 	src->ops->dsi.disable(src, false, true);
 
 	ddata->ulps_enabled = true;
@@ -277,9 +265,6 @@  static int dsicm_exit_ulps(struct panel_drv_data *ddata)
 		goto err2;
 	}
 
-	if (ddata->ext_te_gpio)
-		enable_irq(gpiod_to_irq(ddata->ext_te_gpio));
-
 	dsicm_queue_ulps_work(ddata);
 
 	ddata->ulps_enabled = false;
@@ -290,11 +275,8 @@  static int dsicm_exit_ulps(struct panel_drv_data *ddata)
 	dev_err(&ddata->dsi->dev, "failed to exit ULPS");
 
 	r = dsicm_panel_reset(ddata);
-	if (!r) {
-		if (ddata->ext_te_gpio)
-			enable_irq(gpiod_to_irq(ddata->ext_te_gpio));
+	if (!r)
 		ddata->ulps_enabled = false;
-	}
 
 	dsicm_queue_ulps_work(ddata);
 
@@ -600,7 +582,7 @@  static int dsicm_power_on(struct panel_drv_data *ddata)
 	if (r)
 		goto err;
 
-	r = _dsicm_enable_te(ddata, ddata->te_enabled);
+	r = _dsicm_enable_te(ddata, true);
 	if (r)
 		goto err;
 
@@ -744,43 +726,6 @@  static void dsicm_framedone_cb(int err, void *data)
 	src->ops->dsi.bus_unlock(src);
 }
 
-static irqreturn_t dsicm_te_isr(int irq, void *data)
-{
-	struct panel_drv_data *ddata = data;
-	struct omap_dss_device *src = ddata->src;
-	int old;
-	int r;
-
-	old = atomic_cmpxchg(&ddata->do_update, 1, 0);
-
-	if (old) {
-		cancel_delayed_work(&ddata->te_timeout_work);
-
-		r = src->ops->dsi.update(src, ddata->dsi->channel, dsicm_framedone_cb,
-				ddata);
-		if (r)
-			goto err;
-	}
-
-	return IRQ_HANDLED;
-err:
-	dev_err(&ddata->dsi->dev, "start update failed\n");
-	src->ops->dsi.bus_unlock(src);
-	return IRQ_HANDLED;
-}
-
-static void dsicm_te_timeout_work_callback(struct work_struct *work)
-{
-	struct panel_drv_data *ddata = container_of(work, struct panel_drv_data,
-					te_timeout_work.work);
-	struct omap_dss_device *src = ddata->src;
-
-	dev_err(&ddata->dsi->dev, "TE not received for 250ms!\n");
-
-	atomic_set(&ddata->do_update, 0);
-	src->ops->dsi.bus_unlock(src);
-}
-
 static int dsicm_update(struct omap_dss_device *dssdev,
 				    u16 x, u16 y, u16 w, u16 h)
 {
@@ -808,16 +753,10 @@  static int dsicm_update(struct omap_dss_device *dssdev,
 	if (r)
 		goto err;
 
-	if (ddata->te_enabled && ddata->ext_te_gpio) {
-		schedule_delayed_work(&ddata->te_timeout_work,
-				msecs_to_jiffies(250));
-		atomic_set(&ddata->do_update, 1);
-	} else {
-		r = src->ops->dsi.update(src, ddata->dsi->channel, dsicm_framedone_cb,
-				ddata);
-		if (r)
-			goto err;
-	}
+	r = src->ops->dsi.update(src, ddata->dsi->channel, dsicm_framedone_cb,
+			ddata);
+	if (r)
+		goto err;
 
 	/* note: no bus_unlock here. unlock is src framedone_cb */
 	mutex_unlock(&ddata->lock);
@@ -839,8 +778,7 @@  static int _dsicm_enable_te(struct panel_drv_data *ddata, bool enable)
 	else
 		r = mipi_dsi_dcs_set_tear_off(dsi);
 
-	if (!ddata->ext_te_gpio)
-		src->ops->dsi.enable_te(src, enable);
+	src->ops->dsi.enable_te(src, enable);
 
 	/* possible panel bug */
 	msleep(100);
@@ -933,14 +871,6 @@  static int dsicm_probe_of(struct mipi_dsi_device *dsi)
 		return err;
 	}
 
-	ddata->ext_te_gpio = devm_gpiod_get_optional(&dsi->dev, "te",
-						     GPIOD_IN);
-	if (IS_ERR(ddata->ext_te_gpio)) {
-		err = PTR_ERR(ddata->ext_te_gpio);
-		dev_err(&dsi->dev, "TE gpio request failed: %d", err);
-		return err;
-	}
-
 	err = of_get_display_timing(node, "panel-timing", &timing);
 	if (!err) {
 		videomode_from_timing(&timing, &ddata->vm);
@@ -1023,25 +953,6 @@  static int dsicm_probe(struct mipi_dsi_device *dsi)
 
 	mutex_init(&ddata->lock);
 
-	atomic_set(&ddata->do_update, 0);
-
-	if (ddata->ext_te_gpio) {
-		r = devm_request_irq(dev, gpiod_to_irq(ddata->ext_te_gpio),
-				dsicm_te_isr,
-				IRQF_TRIGGER_RISING,
-				"taal vsync", ddata);
-
-		if (r) {
-			dev_err(dev, "IRQ request failed\n");
-			goto err_reg;
-		}
-
-		INIT_DEFERRABLE_WORK(&ddata->te_timeout_work,
-					dsicm_te_timeout_work_callback);
-
-		dev_dbg(dev, "Using GPIO TE\n");
-	}
-
 	ddata->workqueue = create_singlethread_workqueue("dsicm_wq");
 	if (!ddata->workqueue) {
 		r = -ENOMEM;
diff --git a/drivers/gpu/drm/omapdrm/dss/dsi.c b/drivers/gpu/drm/omapdrm/dss/dsi.c
index 352c16a37476..b26cbedd7e76 100644
--- a/drivers/gpu/drm/omapdrm/dss/dsi.c
+++ b/drivers/gpu/drm/omapdrm/dss/dsi.c
@@ -14,7 +14,9 @@ 
 #include <linux/device.h>
 #include <linux/err.h>
 #include <linux/interrupt.h>
+#include <linux/irq.h>
 #include <linux/delay.h>
+#include <linux/gpio/consumer.h>
 #include <linux/mutex.h>
 #include <linux/module.h>
 #include <linux/semaphore.h>
@@ -368,6 +370,12 @@  struct dsi_data {
 	unsigned int update_bytes;
 #endif
 
+	/* external TE GPIO */
+	struct gpio_desc *te_gpio;
+	int te_irq;
+	struct delayed_work te_timeout_work;
+	atomic_t do_ext_te_update;
+
 	bool te_enabled;
 	bool ulps_enabled;
 
@@ -3826,19 +3834,12 @@  static void dsi_framedone_irq_callback(void *data)
 	dsi_handle_framedone(dsi, 0);
 }
 
-static int dsi_update(struct omap_dss_device *dssdev, int channel,
-		void (*callback)(int, void *), void *data)
+static int _dsi_update(struct dsi_data *dsi)
 {
-	struct dsi_data *dsi = to_dsi_data(dssdev);
 	u16 dw, dh;
 
 	dsi_perf_mark_setup(dsi);
 
-	dsi->update_channel = channel;
-
-	dsi->framedone_callback = callback;
-	dsi->framedone_data = data;
-
 	dw = dsi->vm.hactive;
 	dh = dsi->vm.vactive;
 
@@ -3851,6 +3852,26 @@  static int dsi_update(struct omap_dss_device *dssdev, int channel,
 	return 0;
 }
 
+static int dsi_update(struct omap_dss_device *dssdev, int channel,
+		void (*callback)(int, void *), void *data)
+{
+	struct dsi_data *dsi = to_dsi_data(dssdev);
+
+	dsi->update_channel = channel;
+	dsi->framedone_callback = callback;
+	dsi->framedone_data = data;
+
+	if (dsi->te_enabled && dsi->te_gpio) {
+		schedule_delayed_work(&dsi->te_timeout_work,
+				      msecs_to_jiffies(250));
+		atomic_set(&dsi->do_ext_te_update, 1);
+	} else {
+		_dsi_update(dsi);
+	}
+
+	return 0;
+}
+
 /* Display funcs */
 
 static int dsi_configure_dispc_clocks(struct dsi_data *dsi)
@@ -4094,6 +4115,14 @@  static int dsi_enable_te(struct omap_dss_device *dssdev, bool enable)
 	struct dsi_data *dsi = to_dsi_data(dssdev);
 
 	dsi->te_enabled = enable;
+
+	if (dsi->te_gpio) {
+		if (enable)
+			enable_irq(dsi->te_irq);
+		else
+			disable_irq(dsi->te_irq);
+	}
+
 	return 0;
 }
 
@@ -4770,11 +4799,96 @@  static const struct omap_dss_device_ops dsi_ops = {
 	},
 };
 
+static irqreturn_t omap_dsi_te_irq_handler(int irq, void *dev_id)
+{
+	struct dsi_data *dsi = (struct dsi_data *)dev_id;
+	int old;
+
+	old = atomic_cmpxchg(&dsi->do_ext_te_update, 1, 0);
+	if (old) {
+		cancel_delayed_work(&dsi->te_timeout_work);
+		_dsi_update(dsi);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static void omap_dsi_te_timeout_work_callback(struct work_struct *work)
+{
+	struct dsi_data *dsi =
+		container_of(work, struct dsi_data, te_timeout_work.work);
+	int old;
+
+	old = atomic_cmpxchg(&dsi->do_ext_te_update, 1, 0);
+	if (old) {
+		dev_err(dsi->dev, "TE not received for 250ms!\n");
+		_dsi_update(dsi);
+	}
+}
+
+static int omap_dsi_register_te_irq(struct dsi_data *dsi,
+				    struct mipi_dsi_device *client)
+{
+	int err;
+	int te_irq;
+
+	dsi->te_gpio = gpiod_get(&client->dev, "te-gpios", GPIOD_IN);
+	if (IS_ERR(dsi->te_gpio)) {
+		err = PTR_ERR(dsi->te_gpio);
+
+		if (err == -ENOENT) {
+			dsi->te_gpio = NULL;
+			return 0;
+		}
+
+		dev_err(dsi->dev, "Could not get TE gpio: %d\n", err);
+		return err;
+	}
+
+	te_irq = gpiod_to_irq(dsi->te_gpio);
+	if (te_irq < 0) {
+		gpiod_put(dsi->te_gpio);
+		dsi->te_gpio = NULL;
+		return -EINVAL;
+	}
+
+	dsi->te_irq = te_irq;
+
+	irq_set_status_flags(te_irq, IRQ_NOAUTOEN);
+
+	err = request_threaded_irq(te_irq, NULL, omap_dsi_te_irq_handler,
+				   IRQF_TRIGGER_RISING, "TE", dsi);
+	if (err) {
+		dev_err(dsi->dev, "request irq failed with %d\n", err);
+		gpiod_put(dsi->te_gpio);
+		dsi->te_gpio = NULL;
+		return err;
+	}
+
+	INIT_DEFERRABLE_WORK(&dsi->te_timeout_work,
+			     omap_dsi_te_timeout_work_callback);
+
+	dev_dbg(dsi->dev, "Using GPIO TE\n");
+
+	return 0;
+}
+
+static void omap_dsi_unregister_te_irq(struct dsi_data *dsi)
+{
+	if (dsi->te_gpio) {
+		free_irq(dsi->te_irq, dsi);
+		cancel_delayed_work(&dsi->te_timeout_work);
+		gpiod_put(dsi->te_gpio);
+		dsi->te_gpio = NULL;
+	}
+}
+
 static int omap_dsi_host_attach(struct mipi_dsi_host *host,
 				struct mipi_dsi_device *client)
 {
 	struct dsi_data *dsi = host_to_omap(host);
 	unsigned int channel = client->channel;
+	int r;
 
 	if (channel > 3)
 		return -EINVAL;
@@ -4789,13 +4903,20 @@  static int omap_dsi_host_attach(struct mipi_dsi_host *host,
 		return -EINVAL;
 	}
 
-	dsi->vc[channel].dest = client;
+	atomic_set(&dsi->do_ext_te_update, 0);
 
-	dsi->pix_fmt = client->format;
-	if (client->mode_flags & MIPI_DSI_MODE_VIDEO)
+	if (client->mode_flags & MIPI_DSI_MODE_VIDEO) {
 		dsi->mode = OMAP_DSS_DSI_VIDEO_MODE;
-	else
+	} else {
+		r = omap_dsi_register_te_irq(dsi, client);
+		if (r)
+			return r;
+
 		dsi->mode = OMAP_DSS_DSI_CMD_MODE;
+	}
+
+	dsi->vc[channel].dest = client;
+	dsi->pix_fmt = client->format;
 
 	return 0;
 }
@@ -4812,6 +4933,7 @@  static int omap_dsi_host_detach(struct mipi_dsi_host *host,
 	if (dsi->vc[channel].dest != client)
 		return -EINVAL;
 
+	omap_dsi_unregister_te_irq(dsi);
 	dsi->vc[channel].dest = NULL;
 	return 0;
 }