imx: nitrogen6x: mx6qsabrelite: Add support for DVI monitors

PATCH Superseded Robert Winkler 3 participants 4 messages Jul 22, 2013 – Jul 23, 2013

Subsystems (2)

Arm (32-bit) Arm i.MX and Layerscape

lore

Share & tools

lore

Robert Winkler

author

2013-07-22 20:33 UTC #0 lore raw

PATCH 1/1 imx: nitrogen6x: mx6qsabrelite: Add support for DVI monitors

DVI monitors show up as the RX_SENSE bits in the HDMI_PHY_STAT0 register, ie F0 so we really want to check against F2 to support both HDMI and DVI.

Signed-off-by Robert Winkler <[email protected]>

---
 arch/arm/imx-common/cmd_hdmidet.c             | 3 +--
 arch/arm/include/asm/arch-mx6/mxc_hdmi.h      | 3 +++
 board/boundary/nitrogen6x/nitrogen6x.c        | 2 +-
 board/freescale/mx6qsabrelite/mx6qsabrelite.c | 2 +-
 4 files changed, 6 insertions(+), 4 deletions(-)

arch/arm/imx-common/cmd_hdmidet.c +1 -2

index 794bf50..ce0b26e 100644
@@ -27,8 +27,7 @@
 static int do_hdmidet(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 {
 	struct hdmi_regs *hdmi	= (struct hdmi_regs *)HDMI_ARB_BASE_ADDR;
-	u8 reg = readb(&hdmi->phy_stat0) & HDMI_PHY_HPD;
-	return (reg&HDMI_PHY_HPD) ? 0 : 1;
+	return (readb(&hdmi->phy_stat0) & HDMI_DVI_STAT) ? 0 : 1;
 }
 
 U_BOOT_CMD(hdmidet, 1, 1, do_hdmidet,

arch/arm/include/asm/arch-mx6/mxc_hdmi.h +3 -0

index 9dccb3f..1b55b05 100644
@@ -896,6 +896,9 @@ enum {
 	HDMI_PHY_HPD = 0x02,
 	HDMI_PHY_TX_PHY_LOCK = 0x01,
 
+/* Convenience macro RX_SENSE | HPD */
+	HDMI_DVI_STAT = 0xF2,
+
 /* PHY_I2CM_SLAVE_ADDR field values */
 	HDMI_PHY_I2CM_SLAVE_ADDR_PHY_GEN2 = 0x69,
 	HDMI_PHY_I2CM_SLAVE_ADDR_HEAC_PHY = 0x49,

board/boundary/nitrogen6x/nitrogen6x.c +1 -1

index 8f0f9b8..bb09a6a 100644
@@ -477,7 +477,7 @@ struct display_info_t {
 static int detect_hdmi(struct display_info_t const *dev)
 {
 	struct hdmi_regs *hdmi	= (struct hdmi_regs *)HDMI_ARB_BASE_ADDR;
-	return readb(&hdmi->phy_stat0) & HDMI_PHY_HPD;
+	return readb(&hdmi->phy_stat0) & HDMI_DVI_STAT;
 }
 
 static void enable_hdmi(struct display_info_t const *dev)

board/freescale/mx6qsabrelite/mx6qsabrelite.c +1 -1

index 862bc30..559b082 100644
@@ -449,7 +449,7 @@ struct display_info_t {
 static int detect_hdmi(struct display_info_t const *dev)
 {
 	struct hdmi_regs *hdmi	= (struct hdmi_regs *)HDMI_ARB_BASE_ADDR;
-	return readb(&hdmi->phy_stat0) & HDMI_PHY_HPD;
+	return readb(&hdmi->phy_stat0) & HDMI_DVI_STAT;
 }
 
 static void enable_hdmi(struct display_info_t const *dev)

Signature · 1 line

1.8.3.2

Eric Nelson

2013-07-23 21:43 UTC #1 lore raw

PATCH 1/1 imx: nitrogen6x: mx6qsabrelite: Add support for DVI monitors

Hi all,

Quoted 5 lines · Robert Winkler

DVI monitors show up as the RX_SENSE bits in the HDMI_PHY_STAT0 register, ie F0 so we really want to check against F2 to support both HDMI and DVI.

A little background is probably appropriate for this patch.

Since "the beginning" of usage of the SABRE Lite and Nitrogen6x boards, DVI detection has been somewhat broken.

Some (most) DVI monitors don't produce the "HPD" bit in the PHY_STAT0 register, but do show proper toggling of the RX_SENSE0..3 bits.

Creating a new the bit-mask to include all five bits and modifying the 'hdmidet' command and internal detection routines allows these monitors to function properly in U-Boot.

A related patch to our kernels allows things to work under Linux:

	https://github.com/boundarydevices/linux-imx6/commit/7d8752905c118af9063738a533227de0b2f6ecd4

Quoted 27 lines

Signed-off-by Robert Winkler <[email protected]>

---
  arch/arm/imx-common/cmd_hdmidet.c             | 3 +--
  arch/arm/include/asm/arch-mx6/mxc_hdmi.h      | 3 +++
  board/boundary/nitrogen6x/nitrogen6x.c        | 2 +-
  board/freescale/mx6qsabrelite/mx6qsabrelite.c | 2 +-
  4 files changed, 6 insertions(+), 4 deletions(-)

arch/arm/imx-common/cmd_hdmidet.c +1 -2

index 794bf50..ce0b26e 100644
@@ -27,8 +27,7 @@
  static int do_hdmidet(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
  {
  	struct hdmi_regs *hdmi	= (struct hdmi_regs *)HDMI_ARB_BASE_ADDR;
-	u8 reg = readb(&hdmi->phy_stat0) & HDMI_PHY_HPD;
-	return (reg&HDMI_PHY_HPD) ? 0 : 1;
+	return (readb(&hdmi->phy_stat0) & HDMI_DVI_STAT) ? 0 : 1;
  }
 
  U_BOOT_CMD(hdmidet, 1, 1, do_hdmidet,

arch/arm/include/asm/arch-mx6/mxc_hdmi.h +0 -0

index 9dccb3f..1b55b05 100644

arch/arm/include/asm/arch-mx6/mxc_hdmi.h enum { · line 904

Quoted 9 lines

arch/arm/include/asm/arch-mx6/mxc_hdmi.h +3 -0

@@ -896,6 +896,9 @@ enum {
  	HDMI_PHY_HPD = 0x02,
  	HDMI_PHY_TX_PHY_LOCK = 0x01,
 
+/* Convenience macro RX_SENSE | HPD */
+	HDMI_DVI_STAT = 0xF2,
+
  /* PHY_I2CM_SLAVE_ADDR field values */
  	HDMI_PHY_I2CM_SLAVE_ADDR_PHY_GEN2 = 0x69,
  	HDMI_PHY_I2CM_SLAVE_ADDR_HEAC_PHY = 0x49,

FWIW, we've figured out that the root cause of this is current-limiting by DVI monitors interacts with a resistor (R251) that acts as a voltage divider on the HPD signal on our boards.

We've tested this patch against SABRE SDB boards as well as SABRE Lite and Nitrogen6X.

Regards,

Eric

Fabio Estevam

reviewer

2013-07-23 21:46 UTC #2 lore raw

PATCH 1/1 imx: nitrogen6x: mx6qsabrelite: Add support for DVI monitors

Hi Eric,

Quoted 9 lines · Eric Nelson

A little background is probably appropriate for this patch.
Since "the beginning" of usage of the SABRE Lite and Nitrogen6x boards, DVI detection has been somewhat broken.
Some (most) DVI monitors don't produce the "HPD" bit in the PHY_STAT0 register, but do show proper toggling of the RX_SENSE0..3 bits.

Quoted 8 lines

Creating a new the bit-mask to include all five bits and modifying the 'hdmidet' command and internal detection routines allows these monitors to function properly in U-Boot.
A related patch to our kernels allows things to work under Linux:
https://github.com/boundarydevices/linux-imx6/commit/7d8752905c118af9063738a533227de0b2f6ecd4

Thanks for the excellent explanation and for finding the root cause!

It would be really nice if Robert could send a v2 adding your text into the commit log.

Regards,

Fabio Estevam

Eric Nelson

2013-07-23 22:12 UTC #3 lore raw

PATCH 1/1 imx: nitrogen6x: mx6qsabrelite: Add support for DVI monitors

Hi Fabio,

Quoted 21 lines · Fabio Estevam

Hi Eric,
Quoted 16 lines · Eric Nelson
A little background is probably appropriate for this patch.
Since "the beginning" of usage of the SABRE Lite and Nitrogen6x boards, DVI detection has been somewhat broken.
Some (most) DVI monitors don't produce the "HPD" bit in the PHY_STAT0 register, but do show proper toggling of the RX_SENSE0..3 bits.
Creating a new the bit-mask to include all five bits and modifying the 'hdmidet' command and internal detection routines allows these monitors to function properly in U-Boot.
A related patch to our kernels allows things to work under Linux:

Quoted 7 lines

Quoted 1 line
https://github.com/boundarydevices/linux-imx6/commit/7d8752905c118af9063738a533227de0b2f6ecd4
Thanks for the excellent explanation and for finding the root cause!
It would be really nice if Robert could send a v2 adding your text into the commit log.

Seems to be already done.

Also note that this patch will be smaller with a V3 of your patch collapsing SABRE Lite into the nitrogen6x/ tree (nudge nudge)...

Regards,

Eric