[RESEND PATCH] drm/sun4i: hdmi: Improve compatibility with hpd-less HDMI displays

[RESEND PATCH] drm/sun4i: hdmi: Improve compatibility with hpd-less HDMI displays

[Priit Laes]

From: Priit Laes <priit.laes@paf.com>

Even though HDMI connector features hotplug detect pin (HPD), there
are older devices which do not support it. For these devices fall
back to additional check on I2C bus to probe for EDID data.

One known example is HDMI/DVI display with following edid:

$ xxd -p display.edid
00ffffffffffff0005a1e00301000000150f0103800f05780a0f6ea05748
9a2610474f200000010101010101010101010101010101012a08804520e0
0b1020004000953600000018000000fd0034441a2403000a202020202020
0000001000310a20202020202020202020200000001000002a4030701300
782d1100001e006b

$ edid-decode display.edid
EDID version: 1.3
Manufacturer: AMA Model 3e0 Serial Number 1
Digital display
Maximum image size: 15 cm x 5 cm
Gamma: 2.20
RGB color display
First detailed timing is preferred timing
Display x,y Chromaticity:
  Red:   0.6250, 0.3398
  Green: 0.2841, 0.6044
  Blue:  0.1494, 0.0644
  White: 0.2802, 0.3105

Established timings supported:
  640x480@60Hz 4:3 HorFreq: 31469 Hz Clock: 25.175 MHz
Standard timings supported:
Detailed mode: Clock 20.900 MHz, 149 mm x 54 mm
               640  672  672  709 hborder 0
               480  484  484  491 vborder 0
               -hsync -vsync
               VertFreq: 60 Hz, HorFreq: 29478 Hz
Monitor ranges (GTF): 52-68Hz V, 26-36kHz H, max dotclock 30MHz
Dummy block
Dummy block
Checksum: 0x6b (valid)

Now, current implementation is still flawed, as HDMI uses the
HPD signal to indicate that the source should re-read the EDID
due to change in device capabilities. With current HPD polling
implementation we would most certainly miss those notifications
as one can try just swapping two HDMI monitors really fast.

Proper fix would be skipping the HPD pin detection and relying
on just EDID fetching and acting on its changes.

CC: Russell King <linux@armlinux.org.uk>
Signed-off-by: Priit Laes <priit.laes@paf.com>
---
 drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c | 20 ++++++++++++--------
 1 file changed, 12 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c b/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c
index 416da5376701..44d5d50b7ada 100644
--- a/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c
+++ b/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c
@@ -242,14 +242,18 @@ sun4i_hdmi_connector_detect(struct drm_connector *connector, bool force)
 	struct sun4i_hdmi *hdmi = drm_connector_to_sun4i_hdmi(connector);
 	unsigned long reg;
 
-	if (readl_poll_timeout(hdmi->base + SUN4I_HDMI_HPD_REG, reg,
-			       reg & SUN4I_HDMI_HPD_HIGH,
-			       0, 500000)) {
-		cec_phys_addr_invalidate(hdmi->cec_adap);
-		return connector_status_disconnected;
-	}
-
-	return connector_status_connected;
+	/* TODO: Drop HPD polling and instead keep track of EDID changes */
+	if (!readl_poll_timeout(hdmi->base + SUN4I_HDMI_HPD_REG, reg,
+				reg & SUN4I_HDMI_HPD_HIGH,
+				0, 500000))
+		return connector_status_connected;
+
+	/* Fall back to EDID in case display does not support HPD */
+	if (!IS_ERR(hdmi->i2c) && drm_probe_ddc(hdmi->i2c))
+		return connector_status_connected;
+
+	cec_phys_addr_invalidate(hdmi->cec_adap);
+	return connector_status_disconnected;
 }
 
 static const struct drm_connector_funcs sun4i_hdmi_connector_funcs = {
-- 
2.11.0

Re: [RESEND PATCH] drm/sun4i: hdmi: Improve compatibility with hpd-less HDMI displays

[Luc Verhaegen]

On Mon, Mar 04, 2019 at 03:06:16PM +0200, Priit Laes wrote:
> From: Priit Laes <priit.laes-RQIl3+LPqAc@public.gmane.org>
> 
> Even though HDMI connector features hotplug detect pin (HPD), there
> are older devices which do not support it. For these devices fall
> back to additional check on I2C bus to probe for EDID data.
> 
> One known example is HDMI/DVI display with following edid:
> 
> $ xxd -p display.edid
> 00ffffffffffff0005a1e00301000000150f0103800f05780a0f6ea05748
> 9a2610474f200000010101010101010101010101010101012a08804520e0
> 0b1020004000953600000018000000fd0034441a2403000a202020202020
> 0000001000310a20202020202020202020200000001000002a4030701300
> 782d1100001e006b
> 
> $ edid-decode display.edid
> EDID version: 1.3
> Manufacturer: AMA Model 3e0 Serial Number 1
> Digital display
> Maximum image size: 15 cm x 5 cm
> Gamma: 2.20
> RGB color display
> First detailed timing is preferred timing
> Display x,y Chromaticity:
>   Red:   0.6250, 0.3398
>   Green: 0.2841, 0.6044
>   Blue:  0.1494, 0.0644
>   White: 0.2802, 0.3105
> 
> Established timings supported:
>   640x480@60Hz 4:3 HorFreq: 31469 Hz Clock: 25.175 MHz
> Standard timings supported:
> Detailed mode: Clock 20.900 MHz, 149 mm x 54 mm
>                640  672  672  709 hborder 0
>                480  484  484  491 vborder 0
>                -hsync -vsync
>                VertFreq: 60 Hz, HorFreq: 29478 Hz
> Monitor ranges (GTF): 52-68Hz V, 26-36kHz H, max dotclock 30MHz
> Dummy block
> Dummy block
> Checksum: 0x6b (valid)
> 
> Now, current implementation is still flawed, as HDMI uses the
> HPD signal to indicate that the source should re-read the EDID
> due to change in device capabilities. With current HPD polling
> implementation we would most certainly miss those notifications
> as one can try just swapping two HDMI monitors really fast.
> 
> Proper fix would be skipping the HPD pin detection and relying
> on just EDID fetching and acting on its changes.

HPD has been a hard requirement since DDWG came up with DVI somewhere in 
the late 90s. This monitor is plainly broken, and should not get an 
expensive i2c address polling based workaround at the driver level.

Luc Verhaegen.