Skip to contentSkip to conversation

More Haswell VGA fixes

PATCH Applied Paulo Zanoni 3 participants 13 messages Dec 1, 2012 – Dec 10, 2012

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover More Haswell VGA fixes

From: Paulo Zanoni <[email protected]>

Hi

Here are more Haswell VGA fixes that I would like to see on Kernel 3.8. The first 2 patches are needed to make VGA work after suspend, the 3rd patch is needed to prevent black screen on some specific machines.

These patches are currently on top of drm-intel-fixes + the following 2 patches which are already on the mailing list:
- drm/i915: fix hsw_fdi_link_train "retry" code
- drm/i915: reject modes the LPT FDI receiver can't handle

I'd love to get these 5 patches on Kernel 3.8 since they're all important fixes that prevent black screens.

Thanks,
Paulo

Paulo Zanoni (3):
drm/i915: add support for mPHY destination on intel_sbi_{read,write} drm/i915: add lpt_init_pch_refclk
drm/i915: set the LPT FDI RX polarity reversal bit when needed

 drivers/gpu/drm/i915/i915_drv.c      |   3 +-
 drivers/gpu/drm/i915/i915_drv.h      |   9 +-
 drivers/gpu/drm/i915/i915_reg.h      |  11 +-
 drivers/gpu/drm/i915/intel_crt.c     |   8 ++
 drivers/gpu/drm/i915/intel_ddi.c     |   2 +
 drivers/gpu/drm/i915/intel_display.c | 247 +++++++++++++++++++++++++++++------
 6 files changed, 236 insertions(+), 44 deletions(-)
Signature · 1 line
1.7.11.7
#1 lore raw

PATCH 1/3 drm/i915: add support for mPHY destination on intel_sbi_{read, write}

From: Paulo Zanoni <[email protected]>

This way we should be able to write mPHY registers using the Sideband Interface in the next commit. Also fixed some syntax oddities in the related code.

---
 drivers/gpu/drm/i915/i915_drv.h      |  5 +++
 drivers/gpu/drm/i915/i915_reg.h      |  4 +++
 drivers/gpu/drm/i915/intel_display.c | 65 +++++++++++++++++-------------------
 3 files changed, 40 insertions(+), 34 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +5 -0
index 79589bb..8513e1c 100644
@@ -382,6 +382,11 @@ enum intel_pch {
 	PCH_LPT,	/* Lynxpoint PCH */
 };
 
+enum intel_sbi_destination {
+	SBI_ICLK,
+	SBI_MPHY,
+};
+
 #define QUIRK_PIPEA_FORCE (1<<0)
 #define QUIRK_LVDS_SSC_DISABLE (1<<1)
 #define QUIRK_INVERT_BRIGHTNESS (1<<2)
drivers/gpu/drm/i915/i915_reg.h +4 -0
index 97fbd9d..0760425 100644
@@ -4534,6 +4534,10 @@
 #define SBI_ADDR			0xC6000
 #define SBI_DATA			0xC6004
 #define SBI_CTL_STAT			0xC6008
+#define  SBI_CTL_DEST_ICLK		(0x0<<16)
+#define  SBI_CTL_DEST_MPHY		(0x1<<16)
+#define  SBI_CTL_OP_IORD		(0x2<<8)
+#define  SBI_CTL_OP_IOWR		(0x3<<8)
 #define  SBI_CTL_OP_CRRD		(0x6<<8)
 #define  SBI_CTL_OP_CRWR		(0x7<<8)
 #define  SBI_RESPONSE_FAIL		(0x1<<1)
drivers/gpu/drm/i915/intel_display.c +31 -34

119 lines not on this page · raw message

Signature · 1 line
1.7.11.7
Reviewed-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] drm/i915: add support for mPHY destination on intel_sbi_{read, write}

Quoted 7 lines · Paulo Zanoni

From: Paulo Zanoni <[email protected]>

This way we should be able to write mPHY registers using the Sideband Interface in the next commit. Also fixed some syntax oddities in the related code.

(I'd have put the destination parameter first, talk about bike-shedding!)

Signature · 1 line
Damien
#3 lore raw

PATCH 2/3 drm/i915: add lpt_init_pch_refclk

In reply to Paulo Zanoni

From: Paulo Zanoni <[email protected]>

We need this code to init the PCH SSC refclk and the FDI registers. The BIOS does this too and that's why VGA worked before this patch, until you tried to suspend the machine...

This patch implements the "Sequence to enable CLKOUT_DP for FDI usage and configure PCH FDI/IO" from our documentation.

---
 drivers/gpu/drm/i915/i915_drv.c      |   3 +-
 drivers/gpu/drm/i915/i915_drv.h      |   2 +-
 drivers/gpu/drm/i915/i915_reg.h      |   6 +-
 drivers/gpu/drm/i915/intel_display.c | 182 +++++++++++++++++++++++++++++++++--
 4 files changed, 183 insertions(+), 10 deletions(-)
drivers/gpu/drm/i915/i915_drv.c +1 -2
index a129218..530db83 100644
@@ -554,8 +554,7 @@ static int __i915_drm_thaw(struct drm_device *dev)
 
 	/* KMS EnterVT equivalent */
 	if (drm_core_check_feature(dev, DRIVER_MODESET)) {
-		if (HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev))
-			ironlake_init_pch_refclk(dev);
+		intel_init_pch_refclk(dev);
 
 		mutex_lock(&dev->struct_mutex);
 		dev_priv->mm.suspended = 0;
drivers/gpu/drm/i915/i915_drv.h +1 -1
index 8513e1c..65213bc 100644
@@ -1662,7 +1662,7 @@ extern void intel_modeset_setup_hw_state(struct drm_device *dev,
 extern bool intel_fbc_enabled(struct drm_device *dev);
 extern void intel_disable_fbc(struct drm_device *dev);
 extern bool ironlake_set_drps(struct drm_device *dev, u8 val);
-extern void ironlake_init_pch_refclk(struct drm_device *dev);
+extern void intel_init_pch_refclk(struct drm_device *dev);
 extern void gen6_set_rps(struct drm_device *dev, u8 val);
 extern void intel_detect_pch(struct drm_device *dev);
 extern int intel_trans_dp_port_sel(struct drm_crtc *crtc);
drivers/gpu/drm/i915/i915_reg.h +5 -1
index 0760425..acf768d 100644
@@ -3843,7 +3843,9 @@
 #define  FDI_PHASE_SYNC_EN(pipe) (1<<(FDIA_PHASE_SYNC_SHIFT_EN - ((pipe) * 2)))
 #define  FDI_BC_BIFURCATION_SELECT	(1 << 12)
 #define SOUTH_CHICKEN2		0xc2004
-#define  DPLS_EDP_PPS_FIX_DIS	(1<<0)
+#define  FDI_MPHY_IOSFSB_RESET_STATUS	(1<<13)
+#define  FDI_MPHY_IOSFSB_RESET_CTL	(1<<12)
+#define  DPLS_EDP_PPS_FIX_DIS		(1<<0)
 
 #define _FDI_RXA_CHICKEN         0xc200c
 #define _FDI_RXB_CHICKEN         0xc2010
@@ -4555,10 +4557,12 @@
 #define   SBI_SSCDIVINTPHASE_PROPAGATE		(1<<0)
 #define  SBI_SSCCTL				0x020c
 #define  SBI_SSCCTL6				0x060C
+#define   SBI_SSCCTL_PATHALT			(1<<3)
 #define   SBI_SSCCTL_DISABLE			(1<<0)
 #define  SBI_SSCAUXDIV6				0x0610
 #define   SBI_SSCAUXDIV_FINALDIV2SEL(x)		((x)<<4)
 #define  SBI_DBUFF0				0x2a00
+#define   SBI_DBUFF0_ENABLE			(1<<0)
 
 /* LPT PIXCLK_GATE */
 #define PIXCLK_GATE			0xC6020
drivers/gpu/drm/i915/intel_display.c +176 -6

200 lines not on this page · raw message

Signature · 1 line
1.7.11.7
#4 lore raw

Re: [PATCH 2/3] drm/i915: add lpt_init_pch_refclk

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5003

Quoted 13 lines · Paulo Zanoni

From: Paulo Zanoni <[email protected]>

We need this code to init the PCH SSC refclk and the FDI registers. The BIOS does this too and that's why VGA worked before this patch, until you tried to suspend the machine...

This patch implements the "Sequence to enable CLKOUT_DP for FDI usage and configure PCH FDI/IO" from our documentation.

drivers/gpu/drm/i915/intel_display.c +2 -0
+       if (IS_HASWELL(dev) && (dev->pci_device & 0xFF00) == 0x0C00)
+               is_sdv = true;

Maybe have a IS_SDV() or IS_HSW_SDV() macro like we have a IS_ULT() one? The check should be on the PCH revision but hopefully there's a 1:1 correlation?

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5023

Quoted 3 lines
drivers/gpu/drm/i915/intel_display.c +3 -0
+               if (wait_for_atomic_us(I915_READ(SOUTH_CHICKEN2) &
+                                      FDI_MPHY_IOSFSB_RESET_STATUS, 100))
+                       DRM_ERROR("FDI mPHY resert assert timeout\n");

s/resert/reset/

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5033

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+
+               tmp = I915_READ(SOUTH_CHICKEN2);
+               tmp &= ~FDI_MPHY_IOSFSB_RESET_CTL;
+               I915_WRITE(SOUTH_CHICKEN2, tmp);
+
+               if (wait_for_atomic_us((I915_READ(SOUTH_CHICKEN2) &
+                                       FDI_MPHY_IOSFSB_RESET_STATUS) == 0,
+                                      100))
+                       DRM_ERROR("FDI mPHY reset de-assert timeout\n");
+       }

When either of those two waits error out, we carry on the sequence. We should probably fail and disable the VGA output (one of those error paths that never get to be tested, yey! I don't know how well we support connectors disappearing)?

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5141

Quoted 3 lines
drivers/gpu/drm/i915/intel_display.c +3 -0
+       tmp = intel_sbi_read(dev_priv, SBI_DBUFF0, SBI_ICLK);
+       tmp |= SBI_DBUFF0_ENABLE;
+       intel_sbi_write(dev_priv, SBI_DBUFF0, tmp, SBI_ICLK);

Th ULT path is missing there.

I double checked the programming, looks good. I'll follow up with an amended patch for the 2 first bikesheds, leave the error path alone, and add the ULT path as a separate patch.

Signature · 1 line
Damien
#5 lore raw

Re: [PATCH 2/3] drm/i915: add lpt_init_pch_refclk

Hi

2012/12/8 Damien Lespiau <[email protected]>:

Quoted 10 lines · Paulo Zanoni
Quoted 9 lines

From: Paulo Zanoni <[email protected]>

We need this code to init the PCH SSC refclk and the FDI registers. The BIOS does this too and that's why VGA worked before this patch, until you tried to suspend the machine...

This patch implements the "Sequence to enable CLKOUT_DP for FDI usage and configure PCH FDI/IO" from our documentation.

Quoted 8 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5003

Quoted 4 lines
drivers/gpu/drm/i915/intel_display.c +2 -0
+       if (IS_HASWELL(dev) && (dev->pci_device & 0xFF00) == 0x0C00)
+               is_sdv = true;

Maybe have a IS_SDV() or IS_HSW_SDV() macro like we have a IS_ULT() one? The check should be on the PCH revision but hopefully there's a 1:1 correlation?

We should just remove this code in the future, maybe even now... I did not care too much about creating a macro since this is going to be killed.

Quoted 7 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5023

Quoted 4 lines
drivers/gpu/drm/i915/intel_display.c +3 -0
+               if (wait_for_atomic_us(I915_READ(SOUTH_CHICKEN2) &
+                                      FDI_MPHY_IOSFSB_RESET_STATUS, 100))
+                       DRM_ERROR("FDI mPHY resert assert timeout\n");

s/resert/reset/

Quoted 15 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5033

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+
+               tmp = I915_READ(SOUTH_CHICKEN2);
+               tmp &= ~FDI_MPHY_IOSFSB_RESET_CTL;
+               I915_WRITE(SOUTH_CHICKEN2, tmp);
+
+               if (wait_for_atomic_us((I915_READ(SOUTH_CHICKEN2) &
+                                       FDI_MPHY_IOSFSB_RESET_STATUS) == 0,
+                                      100))
+                       DRM_ERROR("FDI mPHY reset de-assert timeout\n");
+       }

When either of those two waits error out, we carry on the sequence. We should probably fail and disable the VGA output (one of those error paths that never get to be tested, yey! I don't know how well we support connectors disappearing)?

We have this "pattern" in many places: the doc tells us "wait for this bit or wait at least XXXms", then we wait using wait_for and print DRM_ERROR in case we timeout. In theory, even if we hit the timeout, we should be fine since we've already waited for the amount of time the spec tells us to wait... I'm not really sure what's the best thing to do here, but it really seems we're probably never hit the "fail" path.

Quoted 6 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5141

Quoted 4 lines
drivers/gpu/drm/i915/intel_display.c +3 -0
+       tmp = intel_sbi_read(dev_priv, SBI_DBUFF0, SBI_ICLK);
+       tmp |= SBI_DBUFF0_ENABLE;
+       intel_sbi_write(dev_priv, SBI_DBUFF0, tmp, SBI_ICLK);

Th ULT path is missing there.

The ULT will never reach this point, there's not VGA.

Quoted 7 lines

I double checked the programming, looks good. I'll follow up with an amended patch for the 2 first bikesheds, leave the error path alone, and add the ULT path as a separate patch.

--
Damien

Signature · 1 line
Paulo Zanoni
#6 lore raw

Re: [PATCH 2/3] drm/i915: add lpt_init_pch_refclk

Quoted 18 lines · Paulo Zanoni

Hi

2012/12/8 Damien Lespiau <[email protected]>:

Quoted 15 lines · Paulo Zanoni
Quoted 13 lines

From: Paulo Zanoni <[email protected]>

We need this code to init the PCH SSC refclk and the FDI registers. The BIOS does this too and that's why VGA worked before this patch, until you tried to suspend the machine...

This patch implements the "Sequence to enable CLKOUT_DP for FDI usage and configure PCH FDI/IO" from our documentation.

drivers/gpu/drm/i915/intel_display.c +2 -0
+       if (IS_HASWELL(dev) && (dev->pci_device & 0xFF00) == 0x0C00)
+               is_sdv = true;
Quoted 7 lines
Quoted 3 lines

Maybe have a IS_SDV() or IS_HSW_SDV() macro like we have a IS_ULT() one? The check should be on the PCH revision but hopefully there's a 1:1 correlation?

We should just remove this code in the future, maybe even now... I did not care too much about creating a macro since this is going to be killed.

Added a comment to remind us of that.

Quoted 23 lines
Quoted 7 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5023

Quoted 4 lines
drivers/gpu/drm/i915/intel_display.c +3 -0
+               if (wait_for_atomic_us(I915_READ(SOUTH_CHICKEN2) &
+                                      FDI_MPHY_IOSFSB_RESET_STATUS, 100))
+                       DRM_ERROR("FDI mPHY resert assert timeout\n");

s/resert/reset/

Quoted 15 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5033

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+
+               tmp = I915_READ(SOUTH_CHICKEN2);
+               tmp &= ~FDI_MPHY_IOSFSB_RESET_CTL;
+               I915_WRITE(SOUTH_CHICKEN2, tmp);
+
+               if (wait_for_atomic_us((I915_READ(SOUTH_CHICKEN2) &
+                                       FDI_MPHY_IOSFSB_RESET_STATUS) == 0,
+                                      100))
+                       DRM_ERROR("FDI mPHY reset de-assert timeout\n");
+       }

When either of those two waits error out, we carry on the sequence. We should probably fail and disable the VGA output (one of those error paths that never get to be tested, yey! I don't know how well we support connectors disappearing)?

Quoted 7 lines

We have this "pattern" in many places: the doc tells us "wait for this bit or wait at least XXXms", then we wait using wait_for and print DRM_ERROR in case we timeout. In theory, even if we hit the timeout, we should be fine since we've already waited for the amount of time the spec tells us to wait... I'm not really sure what's the best thing to do here, but it really seems we're probably never hit the "fail" path.

Imo this is ok. If the hw is in a non-cooperative mood, there's not much we can do than detect it and hope for the best.

Quoted 7 lines
Quoted 5 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5141

Quoted 3 lines
drivers/gpu/drm/i915/intel_display.c +3 -0
+       tmp = intel_sbi_read(dev_priv, SBI_DBUFF0, SBI_ICLK);
+       tmp |= SBI_DBUFF0_ENABLE;
+       intel_sbi_write(dev_priv, SBI_DBUFF0, tmp, SBI_ICLK);

Th ULT path is missing there.

The ULT will never reach this point, there's not VGA.

Added a comment.

Another bikeshed: Can we have less magic here, or are the docs as opaque as the code here? I've merged it for now, but usually Dave' doesn't like this much magic in the code, so a fixup patch would be nice. Imo just giving the register some names should be good enough.

Cheers, Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#7 lore raw

Re: [PATCH 2/3] drm/i915: add lpt_init_pch_refclk

Quoted 9 lines · Daniel Vetter
Quoted 7 lines
Quoted 5 lines

drivers/gpu/drm/i915/intel_display.c void ironlake_init_pch_refclk(struct drm_device *dev) · line 5141

Quoted 3 lines
drivers/gpu/drm/i915/intel_display.c +3 -0
+       tmp = intel_sbi_read(dev_priv, SBI_DBUFF0, SBI_ICLK);
+       tmp |= SBI_DBUFF0_ENABLE;
+       intel_sbi_write(dev_priv, SBI_DBUFF0, tmp, SBI_ICLK);

Th ULT path is missing there.

The ULT will never reach this point, there's not VGA.

Added a comment.

To go to the bottom of this, I wanted to understand why you can enable the PCH SSC.

It is recommended to use the internal SSC source on ULT platforms. The PCH SSC source is then only used for clock bending. Fortunately the PLL Select bits in SPLL_CTL paper over this difference between the internal and PCH SSC sources, selecting the right one based on a fused bit.

And Paulo does check if we have VGA at the start, so indeed, all is good (I believe)!

Signature · 1 line
Damien
#8 lore raw

Re: [PATCH 2/3] drm/i915: add lpt_init_pch_refclk

In reply to Paulo Zanoni
Quoted 10 lines · Paulo Zanoni

Hi

2012/12/8 Damien Lespiau <[email protected]>:

Quoted 3 lines

Maybe have a IS_SDV() or IS_HSW_SDV() macro like we have a IS_ULT() one? The check should be on the PCH revision but hopefully there's a 1:1 correlation?

We should just remove this code in the future, maybe even now... I did not care too much about creating a macro since this is going to be killed.

On the other hand, it's easy to grep for IS_SDV() rather than other more complex tests. We could even expand the macro to other platforms and have a single way to mark SDV code paths.

Signature · 1 line
Damien
#9 lore raw

PATCH 3/3 drm/i915: set the LPT FDI RX polarity reversal bit when needed

In reply to Paulo Zanoni

From: Paulo Zanoni <[email protected]>

If we fail to set the bit when needed we get some nice FDI link training failures (AKA "black screen on VGA output").

While we don't really know how to properly choose whether we need to set the bit or not (VBT?), just read the initial value set by the BIOS and store it for later usage.

---
 drivers/gpu/drm/i915/i915_drv.h  | 2 ++
 drivers/gpu/drm/i915/i915_reg.h  | 1 +
 drivers/gpu/drm/i915/intel_crt.c | 8 ++++++++
 drivers/gpu/drm/i915/intel_ddi.c | 2 ++
 4 files changed, 13 insertions(+)
drivers/gpu/drm/i915/i915_drv.h +2 -0
index 65213bc..557843d 100644
@@ -915,6 +915,8 @@ typedef struct drm_i915_private {
 	bool hw_contexts_disabled;
 	uint32_t hw_context_size;
 
+	bool fdi_rx_polarity_reversed;
+
 	struct i915_suspend_saved_registers regfile;
 
 	/* Old dri1 support infrastructure, beware the dragons ya fools entering
drivers/gpu/drm/i915/i915_reg.h +1 -0
index acf768d..3f75cfa 100644
@@ -3917,6 +3917,7 @@
 #define  FDI_FS_ERRC_ENABLE		(1<<27)
 #define  FDI_FE_ERRC_ENABLE		(1<<26)
 #define  FDI_DP_PORT_WIDTH_X8           (7<<19)
+#define  FDI_RX_POLARITY_REVERSED_LPT	(1<<16)
 #define  FDI_8BPC                       (0<<16)
 #define  FDI_10BPC                      (1<<16)
 #define  FDI_6BPC                       (2<<16)
drivers/gpu/drm/i915/intel_crt.c +8 -0
index 3084d01..fe20bf7 100644
@@ -798,4 +798,12 @@ void intel_crt_init(struct drm_device *dev)
 	crt->force_hotplug_required = 0;
 
 	dev_priv->hotplug_supported_mask |= CRT_HOTPLUG_INT_STATUS;
+
+	/*
+	 * TODO: find a proper way to discover whether we need to set the
+	 * polarity reversal bit or not, instead of relying on the BIOS.
+	 */
+	if (HAS_PCH_LPT(dev))
+		dev_priv->fdi_rx_polarity_reversed =
+		     !!(I915_READ(_FDI_RXA_CTL) & FDI_RX_POLARITY_REVERSED_LPT);
 }
drivers/gpu/drm/i915/intel_ddi.c +2 -0
index 3264cb4..4bad0f7 100644
@@ -180,6 +180,8 @@ void hsw_fdi_link_train(struct drm_crtc *crtc)
 	/* Enable the PCH Receiver FDI PLL */
 	rx_ctl_val = FDI_RX_PLL_ENABLE | FDI_RX_ENHANCE_FRAME_ENABLE |
 		     ((intel_crtc->fdi_lanes - 1) << 19);
+	if (dev_priv->fdi_rx_polarity_reversed)
+		rx_ctl_val |= FDI_RX_POLARITY_REVERSED_LPT;
 	I915_WRITE(_FDI_RXA_CTL, rx_ctl_val);
 	POSTING_READ(_FDI_RXA_CTL);
 	udelay(220);
Signature · 1 line
1.7.11.7
#10 lore raw

Re: [PATCH 3/3] drm/i915: set the LPT FDI RX polarity reversal bit when needed

Quoted 10 lines · Paulo Zanoni

From: Paulo Zanoni <[email protected]>

If we fail to set the bit when needed we get some nice FDI link training failures (AKA "black screen on VGA output").

While we don't really know how to properly choose whether we need to set the bit or not (VBT?), just read the initial value set by the BIOS and store it for later usage.

Sad that we did not catch that before. It's not just the polarity (support for exchanging the - and + wires on a differential pair), but also the FDI lane reversal (0 becomes 3, 1 becomes 2) setting that we need to preserve. There's a fused bit for the DMI lane reversal and FDI defaults to following that DMI lane reversal configuration, but we also have a bit to override that behaviour in the FDI RX register, we need to preserve this bit as well.

I think this is the same story for DDI_BUF_CTL, DDI ports support lane reversal.

I could not find either where these settings are stored, preserving what the BIOS does for those seems the best option so far.

Signature · 1 line
Damien
#11 lore raw

Re: [PATCH 3/3] drm/i915: set the LPT FDI RX polarity reversal bit when needed

Hi

2012/12/8 Damien Lespiau <[email protected]>:

Quoted 12 lines · Paulo Zanoni
Quoted 10 lines

From: Paulo Zanoni <[email protected]>

If we fail to set the bit when needed we get some nice FDI link training failures (AKA "black screen on VGA output").

While we don't really know how to properly choose whether we need to set the bit or not (VBT?), just read the initial value set by the BIOS and store it for later usage.

Quoted 7 lines

Sad that we did not catch that before. It's not just the polarity (support for exchanging the - and + wires on a differential pair), but also the FDI lane reversal (0 becomes 3, 1 becomes 2) setting that we need to preserve. There's a fused bit for the DMI lane reversal and FDI defaults to following that DMI lane reversal configuration, but we also have a bit to override that behaviour in the FDI RX register, we need to preserve this bit as well.

They're different problems and require different patches. This one fixes the "FDI RX polarity" problem, not "lane reversal".

Quoted 8 lines

I think this is the same story for DDI_BUF_CTL, DDI ports support lane reversal.

I could not find either where these settings are stored, preserving what the BIOS does for those seems the best option so far.

--
Damien

Signature · 1 line
Paulo Zanoni
#12 lore raw

Re: [PATCH 3/3] drm/i915: set the LPT FDI RX polarity reversal bit when needed

Quoted 23 lines · Paulo Zanoni

Hi

2012/12/8 Damien Lespiau <[email protected]>:

Quoted 12 lines · Paulo Zanoni
Quoted 10 lines

From: Paulo Zanoni <[email protected]>

If we fail to set the bit when needed we get some nice FDI link training failures (AKA "black screen on VGA output").

While we don't really know how to properly choose whether we need to set the bit or not (VBT?), just read the initial value set by the BIOS and store it for later usage.

Quoted 7 lines

Sad that we did not catch that before. It's not just the polarity (support for exchanging the - and + wires on a differential pair), but also the FDI lane reversal (0 becomes 3, 1 becomes 2) setting that we need to preserve. There's a fused bit for the DMI lane reversal and FDI defaults to following that DMI lane reversal configuration, but we also have a bit to override that behaviour in the FDI RX register, we need to preserve this bit as well.

Quoted 2 lines

They're different problems and require different patches. This one fixes the "FDI RX polarity" problem, not "lane reversal".

Picked up for -fixes (together with patch 1), thanks for the patch. -Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch