Skip to contentSkip to conversation

drm/i915/BXT: Get pipe conf from the port registers

PATCH In review Ramalingam C 5 participants 28 messages Mar 29, 2016 – Apr 19, 2016

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

PATCH 1/2 drm/i915/BXT: Get pipe conf from the port registers

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers.

The required changes are added for BXT in intel_dsi_get_config (encoder->get_config).

---
Previously reviewed at https://patchwork.freedesktop.org/patch/75301/

 drivers/gpu/drm/i915/i915_reg.h      |    1 +
 drivers/gpu/drm/i915/intel_display.c |   44 ++++++++++++++
 drivers/gpu/drm/i915/intel_dsi.c     |  104 ++++++++++++++++++++++++++++++++++
 3 files changed, 149 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +1 -0
index c839ce9..da3cdef 100644
@@ -8210,6 +8210,7 @@ enum skl_disp_power_wells {
 #define  BXT_PIPE_SELECT_SHIFT				7
 #define  BXT_PIPE_SELECT_MASK				(7 << 7)
 #define  BXT_PIPE_SELECT(pipe)				((pipe) << 7)
+#define  BXT_PORT_TO_PIPE(ctrl)		((ctrl & BXT_PIPE_SELECT_MASK) >> 7)
 
 #define _MIPIA_DATA_ADDRESS		(dev_priv->mipi_mmio_base + 0xb108)
 #define _MIPIC_DATA_ADDRESS		(dev_priv->mipi_mmio_base + 0xb908)
drivers/gpu/drm/i915/intel_display.c +44 -0

61 lines not on this page · raw message

drivers/gpu/drm/i915/intel_dsi.c +104 -0

118 lines not on this page · raw message

Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

PATCH 2/2 drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

 drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
 1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
 			  bool adjust)
 {
 	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
 
 #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
 	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
 		ret = false; \
 	}
 
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
 #define PIPE_CONF_CHECK_M_N(name) \
 	if (!intel_compare_link_m_n(&current_config->name, \
 				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
 	PIPE_CONF_CHECK_I(has_dsi_encoder);
 
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
 
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
 
 #undef PIPE_CONF_CHECK_X
 #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
 #undef PIPE_CONF_CHECK_P
 #undef PIPE_CONF_CHECK_I_ALT
 #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

reproduce:

Hi Ramalingam,

[auto build test ERROR on drm-intel/for-linux-next]
[cannot apply to v4.6-rc1 next-20160329]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]

url:    https://github.com/0day-ci/linux/commits/Ramalingam-C/drm-i915-BXT-Get-pipe-conf-from-the-port-registers/20160330-014708
base:   git://anongit.freedesktop.org/drm-intel for-linux-next
config: i386-randconfig-s1-201613 (attached as .config)
reproduce:
        # save the attached .config to linux build tree
        make ARCH=i386 

All errors (new ones prefixed by >>):

In file included from include/linux/list.h:8:0,

                    from include/linux/dmi.h:4,
                    from drivers/gpu/drm/i915/intel_display.c:27:
   drivers/gpu/drm/i915/intel_display.c: In function 'intel_pipe_config_compare':
>> drivers/gpu/drm/i915/intel_display.c:12612:4: error: implicit declaration of function 'dsi_pixel_format_bpp' [-Werror=implicit-function-declaration]
      (dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
       ^
   include/linux/kernel.h:67:36: note: in definition of macro 'DIV_ROUND_UP'
    #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
                                       ^
   drivers/gpu/drm/i915/intel_display.c:12624:4: note: in expansion of macro 'MAX_BXT_DSI_TIMING_RETRIVAL_ERR'
       MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
       ^
   drivers/gpu/drm/i915/intel_display.c:12636:4: note: in expansion of macro 'BXT_DSI_PIPE_CONF_CHECK_I_RANGE'
       BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
       ^
   drivers/gpu/drm/i915/intel_display.c:12751:2: note: in expansion of macro 'PIPE_CONF_CHECK_I_RANGE'
     PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
     ^

cc1: some warnings being treated as errors

vim +/dsi_pixel_format_bpp +12612 drivers/gpu/drm/i915/intel_display.c

12606 * the converson error, hence we consider the 3 times of error as tolerance.
12607 */
12608
12609 #define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
12610 (intel_dsi == NULL ? 0 : \
12611 DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
12612 (dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
12613 intel_dsi->burst_mode_ratio)))
12614
12615 #define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#3 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

In reply to Ramalingam C
Quoted 77 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

 drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
 1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +30 -0
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
 			  bool adjust)
 {
 	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
 
 #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
 	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
 		ret = false; \
 	}
 
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \

drivers/gpu/drm/i915/intel_display.c intel_pipe_config_compare(struct drm_device *dev, · line 12635

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \

Please drop the platform and encoder type checks here, and move them to a higher level.

drivers/gpu/drm/i915/intel_display.c intel_pipe_config_compare(struct drm_device *dev, · line 12645

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}

I think you should model this after PIPE_CONF_CHECK_CLOCK_FUZZY. Please add a function to do the check, similar to intel_fuzzy_clock_check(), and this will be much easier to read and understand.

Maybe call the macro PIPE_CONF_CHECK_DSI_TIMING_FUZZY or something, since I guess it has to be encoder specific.

Quoted 8 lines
drivers/gpu/drm/i915/intel_display.c +1 -0
+
 #define PIPE_CONF_CHECK_M_N(name) \
 	if (!intel_compare_link_m_n(&current_config->name, \
 				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
 	PIPE_CONF_CHECK_I(has_dsi_encoder);
 
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);

drivers/gpu/drm/i915/intel_display.c intel_pipe_config_compare(struct drm_device *dev, · line 12755

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +5 -5
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);

Please add the platform and encoder type checks at this level for clarity (and as I suggested to drop the checks from the macro itself).

	if (IS_BROXTON(dev_priv) && crtc->config->has_dsi_encoder) {
        	...
        } else {
        	...
        }

BR,
Jani.

Quoted 11 lines
drivers/gpu/drm/i915/intel_display.c +1 -0
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
 
 #undef PIPE_CONF_CHECK_X
 #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
 #undef PIPE_CONF_CHECK_P
 #undef PIPE_CONF_CHECK_I_ALT
 #undef PIPE_CONF_CHECK_FLAGS
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#4 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

In reply to Ramalingam C
Quoted 19 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))
Quoted 7 lines

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Quoted 107 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

 drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
 1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
 			  bool adjust)
 {
 	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
 
 #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
 	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
 		ret = false; \
 	}
 
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
 #define PIPE_CONF_CHECK_M_N(name) \
 	if (!intel_compare_link_m_n(&current_config->name, \
 				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
 	PIPE_CONF_CHECK_I(has_dsi_encoder);
 
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
 
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
 
 #undef PIPE_CONF_CHECK_X
 #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
 #undef PIPE_CONF_CHECK_P
 #undef PIPE_CONF_CHECK_I_ALT
 #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#5 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters. Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario. And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors. So IMHO, even if this is possible, it will look more messy.

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable. Please let me know if i can go ahead in this approach.

Quoted 108 lines
Quoted 108 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

  drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
  1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
  			  bool adjust)
  {
  	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
 
  #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
  	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
  		ret = false; \
  	}
 
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
  #define PIPE_CONF_CHECK_M_N(name) \
  	if (!intel_compare_link_m_n(&current_config->name, \
  				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
  	PIPE_CONF_CHECK_I(has_dsi_encoder);
 
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
 
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
 
  #undef PIPE_CONF_CHECK_X
  #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
  #undef PIPE_CONF_CHECK_P
  #undef PIPE_CONF_CHECK_I_ALT
  #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Thanks,
--Ram

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#6 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 47 lines · Ramalingam C
Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters.
Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from
hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario.
And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards
and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors.
So IMHO, even if this is possible, it will look more messy.

Quoted 3 lines

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable.
Please let me know if i can go ahead in this approach.

Yeah I discussed this some more with Jani on irc. I'd say we should store this adjusted horizontal timings (the ones fudged with burst_mode_ratio, lane_count, dual-link and all these things applied) into crtc_state->base.adjusted_mode. And then ofc also read those values out.

The overall idea of the state verify/compare logic is that we start out with requested state from userspace, then derive the real hw state. And then compare that computed hw state with what's there already. Except for clocks, where there's special reasons, we never go back, since going back requires us to apply a range. This is the only way to guarnatee that "hw has the same exact mode programmed in both cases" iff "intel_crtc_state matches per intel_crtc_config_compare".

state->adjusted_mode is never exposed to userspace, so there's no problem if it's has "strange" values. And we already have pipe_src_h/w to express the logical input rectangle.

The idea is similar to how we set adjusted_mode.flags to what we actually program, instead of trying to make something up that's not perfectly accurate.
-Daniel

Quoted 108 lines
Quoted 103 lines
Quoted 103 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

 drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
 1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
 			  bool adjust)
 {
 	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
 #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
 	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
 		ret = false; \
 	}
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
 #define PIPE_CONF_CHECK_M_N(name) \
 	if (!intel_compare_link_m_n(&current_config->name, \
 				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
 	PIPE_CONF_CHECK_I(has_dsi_encoder);
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
 	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
 #undef PIPE_CONF_CHECK_X
 #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
 #undef PIPE_CONF_CHECK_P
 #undef PIPE_CONF_CHECK_I_ALT
 #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Thanks,
--Ram
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#7 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 63 lines · Daniel Vetter
Quoted 46 lines · Ramalingam C
Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters.
Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from
hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario.
And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards
and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors.
So IMHO, even if this is possible, it will look more messy.

Quoted 3 lines

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable.
Please let me know if i can go ahead in this approach.

Yeah I discussed this some more with Jani on irc. I'd say we should store this adjusted horizontal timings (the ones fudged with burst_mode_ratio, lane_count, dual-link and all these things applied) into crtc_state->base.adjusted_mode. And then ofc also read those values out.

The overall idea of the state verify/compare logic is that we start out with requested state from userspace, then derive the real hw state. And then compare that computed hw state with what's there already. Except for clocks, where there's special reasons, we never go back, since going back requires us to apply a range. This is the only way to guarnatee that "hw has the same exact mode programmed in both cases" iff "intel_crtc_state matches per intel_crtc_config_compare".

Quoted 8 lines

state->adjusted_mode is never exposed to userspace, so there's no problem if it's has "strange" values. And we already have pipe_src_h/w to express the logical input rectangle.

The idea is similar to how we set adjusted_mode.flags to what we actually program, instead of trying to make something up that's not perfectly accurate.
-Daniel

Daniel,

I have tested by adjusting the adjusted_mode in set_dsi_timings() instead of intel_dsi_compute_config(). Reason is if we modify the adjusted mode at intel_dsi_compute_config() itself, then modified value will
be taken as input for set_dsi_timings. Hence the get_config will deviate further. I hope this should be fine with you and Jani.

This will work out, if set_dsi_timings() is called after the dsi_compute_config() on every suspend and resume or modeset. I will verify this on Android once and update.

Please share your view on this, so that can update the patch with corresponding changes.

Quoted 107 lines
Quoted 107 lines
Quoted 103 lines
Quoted 103 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

  drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
  1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
  			  bool adjust)
  {
  	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
  #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
  	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
  		ret = false; \
  	}
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
  #define PIPE_CONF_CHECK_M_N(name) \
  	if (!intel_compare_link_m_n(&current_config->name, \
  				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
  	PIPE_CONF_CHECK_I(has_dsi_encoder);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
  #undef PIPE_CONF_CHECK_X
  #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
  #undef PIPE_CONF_CHECK_P
  #undef PIPE_CONF_CHECK_I_ALT
  #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Thanks,
--Ram
Signature · 7 lines
Thanks,
--Ram

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#8 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 81 lines · Ramalingam C
Quoted 63 lines · Daniel Vetter
Quoted 46 lines · Ramalingam C
Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters.
Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from
hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario.
And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards
and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors.
So IMHO, even if this is possible, it will look more messy.

Quoted 3 lines

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable.
Please let me know if i can go ahead in this approach.

Yeah I discussed this some more with Jani on irc. I'd say we should store this adjusted horizontal timings (the ones fudged with burst_mode_ratio, lane_count, dual-link and all these things applied) into crtc_state->base.adjusted_mode. And then ofc also read those values out.

The overall idea of the state verify/compare logic is that we start out with requested state from userspace, then derive the real hw state. And then compare that computed hw state with what's there already. Except for clocks, where there's special reasons, we never go back, since going back requires us to apply a range. This is the only way to guarnatee that "hw has the same exact mode programmed in both cases" iff "intel_crtc_state matches per intel_crtc_config_compare".

Quoted 8 lines

state->adjusted_mode is never exposed to userspace, so there's no problem if it's has "strange" values. And we already have pipe_src_h/w to express the logical input rectangle.

The idea is similar to how we set adjusted_mode.flags to what we actually program, instead of trying to make something up that's not perfectly accurate.
-Daniel

Daniel,

I have tested by adjusting the adjusted_mode in set_dsi_timings() instead of intel_dsi_compute_config().
Reason is if we modify the adjusted mode at intel_dsi_compute_config() itself, then modified value will
be taken as input for set_dsi_timings. Hence the get_config will deviate further. I hope this should be fine with you and Jani.

Quoted 6 lines

This will work out, if set_dsi_timings() is called after the dsi_compute_config() on every suspend and resume or modeset. I will verify this on Android once and update.

Please share your view on this, so that can update the patch with corresponding changes.

I can't speak for Daniel, but I think his point was to update adjusted mode in ->compute_config() in a way that can be used directly in set_dsi_timings(). Then, it should be possible to read the timings from the hardware, and compare.

BR,
Jani.

Quoted 107 lines
Quoted 107 lines
Quoted 107 lines
Quoted 103 lines
Quoted 103 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

  drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
  1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
  			  bool adjust)
  {
  	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
  #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
  	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
  		ret = false; \
  	}
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
  #define PIPE_CONF_CHECK_M_N(name) \
  	if (!intel_compare_link_m_n(&current_config->name, \
  				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
  	PIPE_CONF_CHECK_I(has_dsi_encoder);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
  #undef PIPE_CONF_CHECK_X
  #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
  #undef PIPE_CONF_CHECK_P
  #undef PIPE_CONF_CHECK_I_ALT
  #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Thanks,
--Ram
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#9 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 86 lines · Jani Nikula
Quoted 85 lines · Ramalingam C
Quoted 63 lines · Daniel Vetter
Quoted 46 lines · Ramalingam C
Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters.
Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from
hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario.
And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards
and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors.
So IMHO, even if this is possible, it will look more messy.

Quoted 3 lines

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable.
Please let me know if i can go ahead in this approach.

Yeah I discussed this some more with Jani on irc. I'd say we should store this adjusted horizontal timings (the ones fudged with burst_mode_ratio, lane_count, dual-link and all these things applied) into crtc_state->base.adjusted_mode. And then ofc also read those values out.

The overall idea of the state verify/compare logic is that we start out with requested state from userspace, then derive the real hw state. And then compare that computed hw state with what's there already. Except for clocks, where there's special reasons, we never go back, since going back requires us to apply a range. This is the only way to guarnatee that "hw has the same exact mode programmed in both cases" iff "intel_crtc_state matches per intel_crtc_config_compare".

Quoted 8 lines

state->adjusted_mode is never exposed to userspace, so there's no problem if it's has "strange" values. And we already have pipe_src_h/w to express the logical input rectangle.

The idea is similar to how we set adjusted_mode.flags to what we actually program, instead of trying to make something up that's not perfectly accurate.
-Daniel

Daniel,

I have tested by adjusting the adjusted_mode in set_dsi_timings() instead of intel_dsi_compute_config().
Reason is if we modify the adjusted mode at intel_dsi_compute_config() itself, then modified value will
be taken as input for set_dsi_timings. Hence the get_config will deviate further. I hope this should be fine with you and Jani.

This will work out, if set_dsi_timings() is called after the dsi_compute_config() on every suspend and resume or modeset. I will verify this on Android once and update.

Quoted 6 lines
Quoted 2 lines

Please share your view on this, so that can update the patch with corresponding changes.

I can't speak for Daniel, but I think his point was to update adjusted mode in ->compute_config() in a way that can be used directly in set_dsi_timings(). Then, it should be possible to read the timings from the hardware, and compare.

No, thats not possible jani. I think i didn't elaborate the problem statement enough.
If you can read the programmed value from the hardware without any error, then there is no need for this patch itself.

Even if we program the modified adjusted mode, timing parameters read from get_config() will not be same as of modified adjusted mode.

In BXT DSI only available hw registers doesn't provide all timing parameters in terms of pixels but txbyteclkhs. adjusted mode has the parameters(start and end of hsync, htotal and hdisplay and others) in terms of pixels. So some conversion involved in programming few parameters (hfp, hsync and hbp) and also in retrieving them.

As discussed above port registers expects hfp, hsync and hbp interms of txbyteclkhs.

Sequence of programing (set_dsi_timings) the dsi port registers:
parameters from mode ---> (calc hfp, hsync and hbp) ---> (adjust for dual link) ----> (conversion of Pixels to txbyteclkhs) ---> Program to Port register

Sequence of get_config():
Read from port register ---> (conversion of txbyteclkhs to Pixels) ---> (adjust for dual link) ---> (recalculate the adjusted mode parameters from hfp, hsync and hbp and other readings)

Here if we assume the input to the set_dsi_timings is X(adjusted mode parameter), output of get_config() will be (X + delta1). Here delta1 is error due to multiple DIV_ROUND_UP() in the conversion of bytes <===> txbyteclkhs.
So as daniel says if you modify the adjusted_mode in compute_config() itself, input to the set_dsi_timings() will become (X + delta1) and the readings from the get_config() will become (X + delta1 + delta2)

And it wouldn't be appropriate to program the hw with modified adjusted mode. This modification is just to match it with the pipe_config read from hw. Hence adjusted mode can be modified after the hw programming only, so the place to do is end of set_dsi_timings().

Hope I explained the situation enough.

Quoted 111 lines

BR,
Jani.

Quoted 106 lines
Quoted 106 lines
Quoted 106 lines
Quoted 102 lines
Quoted 102 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

   drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
   1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
   			  bool adjust)
   {
   	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
   #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
   	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
   		ret = false; \
   	}
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
   #define PIPE_CONF_CHECK_M_N(name) \
   	if (!intel_compare_link_m_n(&current_config->name, \
   				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
   	PIPE_CONF_CHECK_I(has_dsi_encoder);
   	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
   	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
   	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
   #undef PIPE_CONF_CHECK_X
   #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
   #undef PIPE_CONF_CHECK_P
   #undef PIPE_CONF_CHECK_I_ALT
   #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Thanks,
--Ram
Signature · 7 lines
Thanks,
--Ram

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#10 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 100 lines · Ramalingam C
Quoted 86 lines · Jani Nikula
Quoted 85 lines · Ramalingam C
Quoted 63 lines · Daniel Vetter
Quoted 46 lines · Ramalingam C
Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters.
Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from
hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario.
And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards
and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors.
So IMHO, even if this is possible, it will look more messy.

Quoted 3 lines

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable.
Please let me know if i can go ahead in this approach.

Yeah I discussed this some more with Jani on irc. I'd say we should store this adjusted horizontal timings (the ones fudged with burst_mode_ratio, lane_count, dual-link and all these things applied) into crtc_state->base.adjusted_mode. And then ofc also read those values out.

The overall idea of the state verify/compare logic is that we start out with requested state from userspace, then derive the real hw state. And then compare that computed hw state with what's there already. Except for clocks, where there's special reasons, we never go back, since going back requires us to apply a range. This is the only way to guarnatee that "hw has the same exact mode programmed in both cases" iff "intel_crtc_state matches per intel_crtc_config_compare".

Quoted 8 lines

state->adjusted_mode is never exposed to userspace, so there's no problem if it's has "strange" values. And we already have pipe_src_h/w to express the logical input rectangle.

The idea is similar to how we set adjusted_mode.flags to what we actually program, instead of trying to make something up that's not perfectly accurate.
-Daniel

Daniel,

I have tested by adjusting the adjusted_mode in set_dsi_timings() instead of intel_dsi_compute_config().
Reason is if we modify the adjusted mode at intel_dsi_compute_config() itself, then modified value will
be taken as input for set_dsi_timings. Hence the get_config will deviate further. I hope this should be fine with you and Jani.

This will work out, if set_dsi_timings() is called after the dsi_compute_config() on every suspend and resume or modeset. I will verify this on Android once and update.

Quoted 6 lines
Quoted 2 lines

Please share your view on this, so that can update the patch with corresponding changes.

I can't speak for Daniel, but I think his point was to update adjusted mode in ->compute_config() in a way that can be used directly in set_dsi_timings(). Then, it should be possible to read the timings from the hardware, and compare.

No, thats not possible jani. I think i didn't elaborate the problem statement enough.
If you can read the programmed value from the hardware without any error, then there is no need for this patch itself.

Quoted 7 lines

Even if we program the modified adjusted mode, timing parameters read from get_config() will not be same as of modified adjusted mode.

In BXT DSI only available hw registers doesn't provide all timing parameters in terms of pixels but txbyteclkhs.
adjusted mode has the parameters(start and end of hsync, htotal and hdisplay and others) in terms of pixels.

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.
-Daniel

Quoted 148 lines

So some conversion involved in programming few parameters (hfp, hsync and hbp) and also in retrieving them.

As discussed above port registers expects hfp, hsync and hbp interms of txbyteclkhs.

Sequence of programing (set_dsi_timings) the dsi port registers:
parameters from mode ---> (calc hfp, hsync and hbp) ---> (adjust for dual link) ----> (conversion of Pixels to txbyteclkhs) ---> Program to Port register

Sequence of get_config():
Read from port register ---> (conversion of txbyteclkhs to Pixels) ---> (adjust for dual link) ---> (recalculate the adjusted mode parameters from hfp, hsync and hbp and other readings)

Here if we assume the input to the set_dsi_timings is X(adjusted mode parameter), output of get_config() will be (X + delta1). Here delta1 is error due to multiple DIV_ROUND_UP() in the conversion of bytes <===> txbyteclkhs.
So as daniel says if you modify the adjusted_mode in compute_config() itself, input to the set_dsi_timings() will become (X + delta1) and the readings from the get_config() will become (X + delta1 + delta2)

And it wouldn't be appropriate to program the hw with modified adjusted mode. This modification is just to match it with the pipe_config read from hw.
Hence adjusted mode can be modified after the hw programming only, so the place to do is end of set_dsi_timings().

Hope I explained the situation enough.

Quoted 111 lines

BR,
Jani.

Quoted 106 lines
Quoted 106 lines
Quoted 106 lines
Quoted 102 lines
Quoted 102 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

  drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
  1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
  			  bool adjust)
  {
  	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
  #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
  	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
  		ret = false; \
  	}
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
  #define PIPE_CONF_CHECK_M_N(name) \
  	if (!intel_compare_link_m_n(&current_config->name, \
  				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
  	PIPE_CONF_CHECK_I(has_dsi_encoder);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
  #undef PIPE_CONF_CHECK_X
  #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
  #undef PIPE_CONF_CHECK_P
  #undef PIPE_CONF_CHECK_I_ALT
  #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Thanks,
--Ram
Signature · 2 lines
Thanks,
--Ram
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#11 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 104 lines · Daniel Vetter
Quoted 103 lines · Ramalingam C
Quoted 86 lines · Jani Nikula
Quoted 85 lines · Ramalingam C
Quoted 63 lines · Daniel Vetter
Quoted 46 lines · Ramalingam C
Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters.
Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from
hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario.
And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards
and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors.
So IMHO, even if this is possible, it will look more messy.

Quoted 3 lines

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable.
Please let me know if i can go ahead in this approach.

Yeah I discussed this some more with Jani on irc. I'd say we should store this adjusted horizontal timings (the ones fudged with burst_mode_ratio, lane_count, dual-link and all these things applied) into crtc_state->base.adjusted_mode. And then ofc also read those values out.

The overall idea of the state verify/compare logic is that we start out with requested state from userspace, then derive the real hw state. And then compare that computed hw state with what's there already. Except for clocks, where there's special reasons, we never go back, since going back requires us to apply a range. This is the only way to guarnatee that "hw has the same exact mode programmed in both cases" iff "intel_crtc_state matches per intel_crtc_config_compare".

Quoted 8 lines

state->adjusted_mode is never exposed to userspace, so there's no problem if it's has "strange" values. And we already have pipe_src_h/w to express the logical input rectangle.

The idea is similar to how we set adjusted_mode.flags to what we actually program, instead of trying to make something up that's not perfectly accurate.
-Daniel

Daniel,

I have tested by adjusting the adjusted_mode in set_dsi_timings() instead of intel_dsi_compute_config().
Reason is if we modify the adjusted mode at intel_dsi_compute_config() itself, then modified value will
be taken as input for set_dsi_timings. Hence the get_config will deviate further. I hope this should be fine with you and Jani.

This will work out, if set_dsi_timings() is called after the dsi_compute_config() on every suspend and resume or modeset. I will verify this on Android once and update.

Quoted 6 lines
Quoted 2 lines

Please share your view on this, so that can update the patch with corresponding changes.

I can't speak for Daniel, but I think his point was to update adjusted mode in ->compute_config() in a way that can be used directly in set_dsi_timings(). Then, it should be possible to read the timings from the hardware, and compare.

No, thats not possible jani. I think i didn't elaborate the problem statement enough.
If you can read the programmed value from the hardware without any error, then there is no need for this patch itself.

Even if we program the modified adjusted mode, timing parameters read from get_config() will not be same as of modified adjusted mode.

Quoted 7 lines
Quoted 4 lines

In BXT DSI only available hw registers doesn't provide all timing parameters in terms of pixels but txbyteclkhs.
adjusted mode has the parameters(start and end of hsync, htotal and hdisplay and others) in terms of pixels.

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.

I let Ville convince me there would be problems with that. Ville, care to fill in the details?

BR,
Jani.

Quoted 150 lines

-Daniel

Quoted 148 lines

So some conversion involved in programming few parameters (hfp, hsync and hbp) and also in retrieving them.

As discussed above port registers expects hfp, hsync and hbp interms of txbyteclkhs.

Sequence of programing (set_dsi_timings) the dsi port registers:
parameters from mode ---> (calc hfp, hsync and hbp) ---> (adjust for dual link) ----> (conversion of Pixels to txbyteclkhs) ---> Program to Port register

Sequence of get_config():
Read from port register ---> (conversion of txbyteclkhs to Pixels) ---> (adjust for dual link) ---> (recalculate the adjusted mode parameters from hfp, hsync and hbp and other readings)

Here if we assume the input to the set_dsi_timings is X(adjusted mode parameter), output of get_config() will be (X + delta1). Here delta1 is error due to multiple DIV_ROUND_UP() in the conversion of bytes <===> txbyteclkhs.
So as daniel says if you modify the adjusted_mode in compute_config() itself, input to the set_dsi_timings() will become (X + delta1) and the readings from the get_config() will become (X + delta1 + delta2)

And it wouldn't be appropriate to program the hw with modified adjusted mode. This modification is just to match it with the pipe_config read from hw.
Hence adjusted mode can be modified after the hw programming only, so the place to do is end of set_dsi_timings().

Hope I explained the situation enough.

Quoted 111 lines

BR,
Jani.

Quoted 106 lines
Quoted 106 lines
Quoted 106 lines
Quoted 102 lines
Quoted 102 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

  drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
  1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
  			  bool adjust)
  {
  	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
  #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
  	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
  		ret = false; \
  	}
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
  #define PIPE_CONF_CHECK_M_N(name) \
  	if (!intel_compare_link_m_n(&current_config->name, \
  				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
  	PIPE_CONF_CHECK_I(has_dsi_encoder);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
  #undef PIPE_CONF_CHECK_X
  #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
  #undef PIPE_CONF_CHECK_P
  #undef PIPE_CONF_CHECK_I_ALT
  #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Thanks,
--Ram
Signature · 2 lines
Thanks,
--Ram
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#12 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 5 lines · Jani Nikula
Quoted 2 lines

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.

I let Ville convince me there would be problems with that. Ville, care to fill in the details?

If we change them too hard the accurate vblank timestamp stuff will be upset. But then we only need to adjust horizontal timings for dsi, whereas on gen5+ the vblank ts code uses the line counter (i.e. vertical timings) only.

If it's just that it should work, and I don't think we have any other users of the adjusted_mode.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#13 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 7 lines · Daniel Vetter
Quoted 5 lines · Jani Nikula
Quoted 2 lines

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.

I let Ville convince me there would be problems with that. Ville, care to fill in the details?

Quoted 7 lines

If we change them too hard the accurate vblank timestamp stuff will be upset. But then we only need to adjust horizontal timings for dsi, whereas on gen5+ the vblank ts code uses the line counter (i.e. vertical timings) only.

If it's just that it should work, and I don't think we have any other users of the adjusted_mode.

Ok, I was wrong and we obviously need the right dotclock to compute linedur_ns correctly in drm_calc_timestamping_constants(). So either we adjust the dotclock of adjusted_mode too (imo makes most sense), or we need yet another mode somewhere and use that for dsi cross checking (real ugly imo). More I missed?
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#14 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 11 lines · Daniel Vetter
Quoted 10 lines · Daniel Vetter
Quoted 4 lines · Jani Nikula
Quoted 2 lines

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.

I let Ville convince me there would be problems with that. Ville, care to fill in the details?

If we change them too hard the accurate vblank timestamp stuff will be upset. But then we only need to adjust horizontal timings for dsi, whereas on gen5+ the vblank ts code uses the line counter (i.e. vertical timings) only.

Quoted 7 lines
Quoted 2 lines

If it's just that it should work, and I don't think we have any other users of the adjusted_mode.

Ok, I was wrong and we obviously need the right dotclock to compute linedur_ns correctly in drm_calc_timestamping_constants(). So either we adjust the dotclock of adjusted_mode too (imo makes most sense), or we need yet another mode somewhere and use that for dsi cross checking (real ugly imo). More I missed?

Another point to be considered:
And we program hsync, hfp and hbp in terms of txbyteclkhs to port register, which are not part of adjusted_mode. So for BXT DSI, we have to store them interms of txbyteclkhs and compare with hsync, hfp and hbp read from HW??

Quoted 1 line

-Daniel

Signature · 7 lines
Thanks,
--Ram

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#15 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 20 lines · Ramalingam C
Quoted 11 lines · Daniel Vetter
Quoted 10 lines · Daniel Vetter
Quoted 4 lines · Jani Nikula
Quoted 2 lines

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.

I let Ville convince me there would be problems with that. Ville, care to fill in the details?

If we change them too hard the accurate vblank timestamp stuff will be upset. But then we only need to adjust horizontal timings for dsi, whereas on gen5+ the vblank ts code uses the line counter (i.e. vertical timings) only.

Quoted 7 lines
Quoted 2 lines

If it's just that it should work, and I don't think we have any other users of the adjusted_mode.

Ok, I was wrong and we obviously need the right dotclock to compute linedur_ns correctly in drm_calc_timestamping_constants(). So either we adjust the dotclock of adjusted_mode too (imo makes most sense), or we need yet another mode somewhere and use that for dsi cross checking (real ugly imo). More I missed?

Quoted 5 lines

Another point to be considered:
And we program hsync, hfp and hbp in terms of txbyteclkhs to port register, which are not part of adjusted_mode.
So for BXT DSI, we have to store them interms of txbyteclkhs and compare with hsync, hfp and hbp read from HW??

Yeah that's my idea. Plus we should probably store txbyteclkhs somewhere, too. Or at least we need to adjust the clock in adjusted_mode to match txbyteclkhs, otherwise the vblank ts code goes off the rails. -Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#16 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 20 lines · Daniel Vetter
Quoted 19 lines · Ramalingam C
Quoted 18 lines · Daniel Vetter
Quoted 5 lines · Daniel Vetter
Quoted 4 lines · Jani Nikula
Quoted 2 lines

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.

I let Ville convince me there would be problems with that. Ville, care to fill in the details?

Quoted 7 lines

If we change them too hard the accurate vblank timestamp stuff will be upset. But then we only need to adjust horizontal timings for dsi, whereas on gen5+ the vblank ts code uses the line counter (i.e. vertical timings) only.

If it's just that it should work, and I don't think we have any other users of the adjusted_mode.

Ok, I was wrong and we obviously need the right dotclock to compute linedur_ns correctly in drm_calc_timestamping_constants(). So either we adjust the dotclock of adjusted_mode too (imo makes most sense), or we need yet another mode somewhere and use that for dsi cross checking (real ugly imo). More I missed?

Quoted 8 lines
Quoted 5 lines

Another point to be considered:
And we program hsync, hfp and hbp in terms of txbyteclkhs to port register, which are not part of adjusted_mode.
So for BXT DSI, we have to store them interms of txbyteclkhs and compare with hsync, hfp and hbp read from HW??

Yeah that's my idea. Plus we should probably store txbyteclkhs somewhere, too. Or at least we need to adjust the clock in adjusted_mode to match txbyteclkhs, otherwise the vblank ts code goes off the rails.

Ok. I will post a RFC for this. But for that we need to finalize few points.

1. We need to add three more variables in pipe_config for caching(SW and HW State) the hfp, hsync and hbp.
2. On BXT DSI, We will compare the hdisplay, hfp, hsync and hbp only on pipe_config_compare not all horizontal timing param of adjusted mode.
3. So we dont need to recalculate all horizontal timing parameters of adjusted_mode from port register at get_config(). Should we fill them too (No harm Though)?

Please clarify if we are fine with above points.

-Ram

Quoted 1 line

-Daniel

Signature · 7 lines
Thanks,
--Ram

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#17 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

Quoted 30 lines · Ramalingam C
Quoted 20 lines · Daniel Vetter
Quoted 19 lines · Ramalingam C
Quoted 18 lines · Daniel Vetter
Quoted 5 lines · Daniel Vetter
Quoted 4 lines · Jani Nikula
Quoted 2 lines

Then fix adjusted_mode to have the timings in terms of txbyteclkhs already. Problem solved.

I let Ville convince me there would be problems with that. Ville, care to fill in the details?

Quoted 7 lines

If we change them too hard the accurate vblank timestamp stuff will be upset. But then we only need to adjust horizontal timings for dsi, whereas on gen5+ the vblank ts code uses the line counter (i.e. vertical timings) only.

If it's just that it should work, and I don't think we have any other users of the adjusted_mode.

Ok, I was wrong and we obviously need the right dotclock to compute linedur_ns correctly in drm_calc_timestamping_constants(). So either we adjust the dotclock of adjusted_mode too (imo makes most sense), or we need yet another mode somewhere and use that for dsi cross checking (real ugly imo). More I missed?

Quoted 8 lines
Quoted 5 lines

Another point to be considered:
And we program hsync, hfp and hbp in terms of txbyteclkhs to port register, which are not part of adjusted_mode.
So for BXT DSI, we have to store them interms of txbyteclkhs and compare with hsync, hfp and hbp read from HW??

Yeah that's my idea. Plus we should probably store txbyteclkhs somewhere, too. Or at least we need to adjust the clock in adjusted_mode to match txbyteclkhs, otherwise the vblank ts code goes off the rails.

Quoted 7 lines

Ok. I will post a RFC for this. But for that we need to finalize few points.

1. We need to add three more variables in pipe_config for caching(SW and HW State) the hfp, hsync and hbp.
2. On BXT DSI, We will compare the hdisplay, hfp, hsync and hbp only on pipe_config_compare not all horizontal timing param of adjusted mode.
3. So we dont need to recalculate all horizontal timing parameters of adjusted_mode from port register at get_config(). Should we fill them too (No harm Though)?

Please clarify if we are fine with above points.

Hi,

Based on the IRC discussion, we decided to nullify the round_up error in the dsi_get_config() itself. I have taken the AR. And the RFC is submitted for review at
https://lists.freedesktop.org/archives/intel-gfx/2016-April/092946.html
https://lists.freedesktop.org/archives/intel-gfx/2016-April/092947.html

Please review the same. Thanks

--Ram

Quoted 4 lines

-Ram

Quoted 1 line

-Daniel

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#18 lore raw

Re: [PATCH 2/2] drm/i915/BXT: Tolerance at BXT DSI pipe_config comparison

In reply to Jani Nikula
Quoted 86 lines · Jani Nikula
Quoted 85 lines · Ramalingam C
Quoted 63 lines · Daniel Vetter
Quoted 46 lines · Ramalingam C
Quoted 26 lines · Daniel Vetter
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers. But mode timing parameters are progammed to port registers interms of byteclocks.

The formula used to convert the pixels interms of byteclk is
DIV_ROUND_UP(DIV_ROUND_UP(pixels * bpp * burst_mode_ratio,
8 * 100), lane_count);

So we retrieve them, interms of pixels as

	DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
					(bpp * burst_mode_ratio));

Due to the multiple DIV_ROUND_UP in both formulas we get the worst case delta in the retrieved PIPE's timing parameter as below

	DIV_ROUND_UP((8 * intel_dsi->lane_count * 100),
		(dsi_pixel_format_bpp(intel_dsi->pixel_format) *
			intel_dsi->burst_mode_ratio)))

This converson of byteclk to pixel is required for hsync, hfp and hbp. Which intern impacts horrizontal timing parameters. At worst case to get htotal all there parameters are added with hactive. Hence delta will be 3 times of above formula. Hence this value is considered as tolerance for pipe_config comparison, in case of BXT DSI.

Quoted 7 lines
Quoted 1 line

This is the wrong way round imo, better would be to adjust the adjusted mode in the bxt dsi compute_config function to match the hw granularity. Stuff _really_ should match exactly, the fuzzy clock matching is mostly because our clock cod is a mess, and we can't/don't properly forward-compuate the actual clock timings we program into the hardware. -Daniel

Daniel, I got your point. But the problem will be that difficulty(even if possible) in adjusting the adjusted mode parameters.
Reason is we are not programing the mode parameter as such. We will derive the hfp, hsync and hbp from
hsync_start, hsync_end, hdisplay and htotal. These will be adjusted(divided by 2) for dual link scenario.
And then resultant will go into the conversion as mentioned in the commit message (two DIV_ROUND_UP onwards
and one DIV_ROUND_UP backwards). For this we have to make the parameter divisible by three different factors.
So IMHO, even if this is possible, it will look more messy.

Quoted 3 lines

Predicting the max error and tolerating it in pipe_config_compare will be the straight forward and more reasonable.
Please let me know if i can go ahead in this approach.

Yeah I discussed this some more with Jani on irc. I'd say we should store this adjusted horizontal timings (the ones fudged with burst_mode_ratio, lane_count, dual-link and all these things applied) into crtc_state->base.adjusted_mode. And then ofc also read those values out.

The overall idea of the state verify/compare logic is that we start out with requested state from userspace, then derive the real hw state. And then compare that computed hw state with what's there already. Except for clocks, where there's special reasons, we never go back, since going back requires us to apply a range. This is the only way to guarnatee that "hw has the same exact mode programmed in both cases" iff "intel_crtc_state matches per intel_crtc_config_compare".

Quoted 8 lines

state->adjusted_mode is never exposed to userspace, so there's no problem if it's has "strange" values. And we already have pipe_src_h/w to express the logical input rectangle.

The idea is similar to how we set adjusted_mode.flags to what we actually program, instead of trying to make something up that's not perfectly accurate.
-Daniel

Daniel,

I have tested by adjusting the adjusted_mode in set_dsi_timings() instead of intel_dsi_compute_config().
Reason is if we modify the adjusted mode at intel_dsi_compute_config() itself, then modified value will
be taken as input for set_dsi_timings. Hence the get_config will deviate further. I hope this should be fine with you and Jani.

This will work out, if set_dsi_timings() is called after the dsi_compute_config() on every suspend and resume or modeset. I will verify this on Android once and update.

Quoted 7 lines
Quoted 2 lines

Please share your view on this, so that can update the patch with corresponding changes.

I can't speak for Daniel, but I think his point was to update adjusted mode in ->compute_config() in a way that can be used directly in set_dsi_timings(). Then, it should be possible to read the timings from the hardware, and compare.

Yup, that's the idea. And the idea furthermore is to store _exactly_ the timings we store in the dsi controller (including clock fudging and whatever) and _not_ try to compute things backwards. The idea behind the hw state checker is that you take the requested mode and compute forward until you are at the values you can program into the hw. Currently what we have in ->adjusted_mode is not actually what we program into the hw, so not yet good enough.

Then the hw state verifier only compares things 1:1. Clocks are a bit an exception, for special reasons. But for anything else we shouldn't ever try to reconstruct the original requested mode from userspace, due to aliasing issues and rounding problems.

This means that for dsi ->adjusted_mode will have even less to do with the requested mode from userspace, but that's no issue. We only use that internally to track the actual hw state.
-Daniel

Quoted 115 lines

BR,
Jani.

Quoted 107 lines
Quoted 107 lines
Quoted 107 lines
Quoted 103 lines
Quoted 103 lines

---
Reviewed at https://lists.freedesktop.org/archives/intel-gfx/2016-March/089548.html

  drivers/gpu/drm/i915/intel_display.c |   62 +++++++++++++++++++++++++++++++---
  1 file changed, 57 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_display.c +57 -5
index c0627d6..282f036 100644
@@ -12557,6 +12557,9 @@ intel_pipe_config_compare(struct drm_device *dev,
  			  bool adjust)
  {
  	bool ret = true;
+	struct intel_crtc *crtc = to_intel_crtc(current_config->base.crtc);
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi = NULL;
  #define INTEL_ERR_OR_DBG_KMS(fmt, ...) \
  	do { \
@@ -12593,6 +12596,54 @@ intel_pipe_config_compare(struct drm_device *dev,
  		ret = false; \
  	}
+/*
+ * In case of BXT DSI, HW pipe_config will be retrieved from the port's timing
+ * configuration. This retrival includes some errors due to the DIV_ROUND_UP.
+ * So we are considering the max possible error at the comparison.
+ */
+/*
+ * htotal = hactive + hfp + hsync + hbp. Here last three lements might have
+ * the converson error, hence we consider the 3 times of error as tolerance.
+ */
+
+#define MAX_BXT_DSI_TIMING_RETRIVAL_ERR \
+		(intel_dsi == NULL ? 0 : \
+		DIV_ROUND_UP((3 * 8 * intel_dsi->lane_count * 100), \
+		(dsi_pixel_format_bpp(intel_dsi->pixel_format) * \
+			intel_dsi->burst_mode_ratio)))
+
+#define BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) { \
+	for_each_encoder_on_crtc(dev, &crtc->base, \
+					intel_encoder) { \
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) { \
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base); \
+		} \
+	} \
+	if (!(current_config->name < pipe_config->name && \
+		current_config->name >= (pipe_config->name - \
+			MAX_BXT_DSI_TIMING_RETRIVAL_ERR))) { \
+		INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+		  "(expected %i, found %i(Err tolerance considered))\n", \
+		  current_config->name, \
+		  pipe_config->name); \
+		ret = false; \
+	} \
+}
+
+#define PIPE_CONF_CHECK_I_RANGE(name) { \
+	if (current_config->name != pipe_config->name) { \
+		if (IS_BROXTON(dev) && crtc->config->has_dsi_encoder) { \
+			BXT_DSI_PIPE_CONF_CHECK_I_RANGE(name) \
+		} else { \
+			INTEL_ERR_OR_DBG_KMS("mismatch in " #name " " \
+			  "(expected %i, found %i)\n", \
+			  current_config->name, \
+			  pipe_config->name); \
+			ret = false; \
+		} \
+	} \
+}
+
  #define PIPE_CONF_CHECK_M_N(name) \
  	if (!intel_compare_link_m_n(&current_config->name, \
  				    &pipe_config->name,\
@@ -12697,11 +12748,11 @@ intel_pipe_config_compare(struct drm_device *dev,
  	PIPE_CONF_CHECK_I(has_dsi_encoder);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hdisplay);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_htotal);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hblank_end);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_start);
-	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_hsync_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_htotal);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hblank_end);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_start);
+	PIPE_CONF_CHECK_I_RANGE(base.adjusted_mode.crtc_hsync_end);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vdisplay);
  	PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vtotal);
@@ -12779,6 +12830,7 @@ intel_pipe_config_compare(struct drm_device *dev,
  #undef PIPE_CONF_CHECK_X
  #undef PIPE_CONF_CHECK_I
+#undef PIPE_CONF_CHECK_I_RANGE
  #undef PIPE_CONF_CHECK_P
  #undef PIPE_CONF_CHECK_I_ALT
  #undef PIPE_CONF_CHECK_FLAGS
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Thanks,
--Ram
Signature · 1 line
Jani Nikula, Intel Open Source Technology Center
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#19 lore raw

✗ Fi.CI.BAT: failure for series starting with [1/2] drm/i915/BXT: Get pipe conf from the port registers

In reply to Ramalingam C
== Series Details ==

== Series Details ==

Series: series starting with [1/2] drm/i915/BXT: Get pipe conf from the port registers
URL : https://patchwork.freedesktop.org/series/5006/
State : failure

== Summary ==

  LD      drivers/video/built-in.o
  CC      drivers/usb/host/xhci-trace.o
  CC [M]  drivers/net/ethernet/intel/e1000e/nvm.o
  CC [M]  drivers/net/ethernet/intel/e1000e/phy.o
  CC      drivers/usb/host/xhci-pci.o
  CC [M]  drivers/net/ethernet/intel/e1000e/param.o
  CC [M]  drivers/net/ethernet/intel/e1000e/ethtool.o
  CC [M]  drivers/net/ethernet/intel/e1000e/netdev.o
  LD [M]  drivers/net/ethernet/intel/e1000/e1000.o
  CC [M]  drivers/net/ethernet/intel/e1000e/ptp.o
  LD [M]  drivers/net/ethernet/intel/igbvf/igbvf.o
  LD      drivers/usb/host/xhci-hcd.o
  LD      drivers/usb/host/built-in.o
  LD      drivers/usb/built-in.o
  LD [M]  drivers/net/ethernet/intel/igb/igb.o

cc1: some warnings being treated as errors
scripts/Makefile.build:291: recipe for target 'drivers/gpu/drm/i915/intel_display.o' failed make[4]: *** [drivers/gpu/drm/i915/intel_display.o] Error 1 scripts/Makefile.build:440: recipe for target 'drivers/gpu/drm/i915' failed make[3]: *** [drivers/gpu/drm/i915] Error 2
scripts/Makefile.build:440: recipe for target 'drivers/gpu/drm' failed make[2]: *** [drivers/gpu/drm] Error 2
scripts/Makefile.build:440: recipe for target 'drivers/gpu' failed
make[1]: *** [drivers/gpu] Error 2
make[1]: *** Waiting for unfinished jobs....

  LD [M]  drivers/net/ethernet/intel/e1000e/e1000e.o
  LD      drivers/net/ethernet/built-in.o
  LD      drivers/net/built-in.o

Makefile:962: recipe for target 'drivers' failed
make: *** [drivers] Error 2

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#20 lore raw
In reply to Ramalingam C
Quoted 83 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers.

The required changes are added for BXT in intel_dsi_get_config (encoder->get_config).

---
Previously reviewed at https://patchwork.freedesktop.org/patch/75301/

 drivers/gpu/drm/i915/i915_reg.h      |    1 +
 drivers/gpu/drm/i915/intel_display.c |   44 ++++++++++++++
 drivers/gpu/drm/i915/intel_dsi.c     |  104 ++++++++++++++++++++++++++++++++++
 3 files changed, 149 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +1 -0
index c839ce9..da3cdef 100644
@@ -8210,6 +8210,7 @@ enum skl_disp_power_wells {
 #define  BXT_PIPE_SELECT_SHIFT				7
 #define  BXT_PIPE_SELECT_MASK				(7 << 7)
 #define  BXT_PIPE_SELECT(pipe)				((pipe) << 7)
+#define  BXT_PORT_TO_PIPE(ctrl)		((ctrl & BXT_PIPE_SELECT_MASK) >> 7)
 
 #define _MIPIA_DATA_ADDRESS		(dev_priv->mipi_mmio_base + 0xb108)
 #define _MIPIC_DATA_ADDRESS		(dev_priv->mipi_mmio_base + 0xb908)
drivers/gpu/drm/i915/intel_display.c +34 -0
index 29aa64b..c0627d6 100644
@@ -9941,11 +9941,40 @@ static void haswell_get_ddi_port_state(struct intel_crtc *crtc,
 	}
 }
 
+struct intel_encoder *bxt_get_dsi_encoder_for_crtc(struct intel_crtc *crtc,
+					struct intel_crtc_state *pipe_config)
+{
+	struct drm_device *dev = crtc->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi;
+	enum port port;
+	u32 tmp;
+
+	for_each_intel_encoder(dev, intel_encoder) {
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) {
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base);
+			for_each_dsi_port(port, intel_dsi->ports) {
+				if (!(I915_READ(BXT_MIPI_PORT_CTRL(port)) &
+								DPI_ENABLE))
+					break;
+
+				tmp = I915_READ(MIPI_CTRL(port));
+				if ((tmp & BXT_PIPE_SELECT_MASK) ==
+						BXT_PIPE_SELECT(crtc->pipe))
+					return intel_encoder;
+			}
+		}
+	}
+	return NULL;
+}
+
 static bool haswell_get_pipe_config(struct intel_crtc *crtc,
 				    struct intel_crtc_state *pipe_config)
 {
 	struct drm_device *dev = crtc->base.dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_encoder *intel_encoder, *attached_encoder = NULL;
 	enum intel_display_power_domain power_domain;
 	unsigned long power_domain_mask;
 	bool active;
@@ -9965,6 +9994,21 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc,
 		WARN_ON(active && pipe_config->has_dsi_encoder);
 		if (pipe_config->has_dsi_encoder)
 			active = true;
+
+		for_each_encoder_on_crtc(dev, &crtc->base, intel_encoder)
+			attached_encoder = intel_encoder;
+
+		/*

drivers/gpu/drm/i915/intel_display.c static bool haswell_get_pipe_config(struct intel_crtc *crtc, · line 10011

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+		 * attached_encoder will be NULL, if there is no modeset from
+		 * the kernel bootup.
+		 */
+		if (!attached_encoder && pipe_config->has_dsi_encoder)
+			attached_encoder =
+				bxt_get_dsi_encoder_for_crtc(crtc, pipe_config);
+
+		if (attached_encoder && attached_encoder->get_config)
+			attached_encoder->get_config(attached_encoder,
+								pipe_config);

No, you must not add a new call to the encoder->get_config() hook.

haswell_get_pipe_config() is called through the dev_priv->display.get_pipe_config() function pointer. This happens in check_crtc_state() and intel_modeset_readout_hw_state(). In both places, encoder->get_config() is called afterwards, if encoder->get_hw_state() returns true for the encoder.

The infrastructure is there, you only need to update DSI ->get_config().

Quoted 28 lines
drivers/gpu/drm/i915/intel_display.c +0 -0
 	}
 
 	if (!active)
drivers/gpu/drm/i915/intel_dsi.c +17 -0
index 0de74e1..69a801e 100644
@@ -740,14 +740,118 @@ out_put_power:
 	return active;
 }
 
+/* return pixels equvalent to txbyteclkhs */
+static u16 pixels_from_txbyteclkhs(u16 clk_hs, int bpp, int lane_count,
+		       u16 burst_mode_ratio)
+{
+	return DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
+						(bpp * burst_mode_ratio));
+}
+
+static void bxt_dsi_get_pipe_config(struct intel_encoder *encoder,
+				 struct intel_crtc_state *pipe_config)
+{
+	struct drm_device *dev = encoder->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_display_mode *adjusted_mode =
+					&pipe_config->base.adjusted_mode;
+	struct intel_dsi *intel_dsi =
+				enc_to_intel_dsi(&encoder->base);

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 769

Quoted 10 lines
drivers/gpu/drm/i915/intel_dsi.c +10 -0
+	unsigned int bpp = mipi_dsi_pixel_format_to_bpp(intel_dsi->pixel_format);
+	unsigned int lane_count = intel_dsi->lane_count;
+	enum port port;
+	enum pipe dsi_pipe;
+	u16 hactive, hfp, hsync, hbp, vfp, vsync, vbp;
+	uint32_t tmp;
+
+	for_each_dsi_port(port, intel_dsi->ports) {
+		if (!(I915_READ(BXT_MIPI_PORT_CTRL(port)) & DPI_ENABLE))
+			continue;

Please use the loop just for figuring out the port to use for reading the registers. After that, you can do the rest at a lower indentation level, and it's clearer the rest only gets done once, for one port.

Quoted 45 lines
drivers/gpu/drm/i915/intel_dsi.c +45 -0
+
+		/* In terms of pixels */
+		adjusted_mode->crtc_hdisplay =
+					I915_READ(BXT_MIPI_TRANS_HACTIVE(port));
+		adjusted_mode->crtc_vdisplay =
+					I915_READ(BXT_MIPI_TRANS_VACTIVE(port));
+		adjusted_mode->crtc_vtotal =
+					I915_READ(BXT_MIPI_TRANS_VTOTAL(port));
+
+		hactive = adjusted_mode->crtc_hdisplay;
+		hfp = I915_READ(MIPI_HFP_COUNT(port));
+
+		/*
+		 * meaningful for video mode non-burst sync pulse mode only,
+		 * can be zero for non-burst sync events and burst modes
+		 */
+		hsync = I915_READ(MIPI_HSYNC_PADDING_COUNT(port));
+		hbp = I915_READ(MIPI_HBP_COUNT(port));
+
+		/* horizontal values are in terms of high speed byte clock */
+		hfp = pixels_from_txbyteclkhs(hfp, bpp, lane_count,
+						intel_dsi->burst_mode_ratio);
+		hsync = pixels_from_txbyteclkhs(hsync, bpp, lane_count,
+						intel_dsi->burst_mode_ratio);
+		hbp = pixels_from_txbyteclkhs(hbp, bpp, lane_count,
+						intel_dsi->burst_mode_ratio);
+
+		if (intel_dsi->dual_link) {
+			hfp *= 2;
+			hsync *= 2;
+			hbp *= 2;
+		}
+
+		/* vertical values are in terms of lines */
+		vfp = I915_READ(MIPI_VFP_COUNT(port));
+		vsync = I915_READ(MIPI_VSYNC_PADDING_COUNT(port));
+		vbp = I915_READ(MIPI_VBP_COUNT(port));
+
+		adjusted_mode->crtc_htotal = hactive + hfp + hsync + hbp;
+		adjusted_mode->crtc_hsync_start =
+					hfp + adjusted_mode->crtc_hdisplay;
+		adjusted_mode->crtc_hsync_end =
+					hsync + adjusted_mode->crtc_hsync_start;
+		adjusted_mode->crtc_hblank_start = adjusted_mode->crtc_hdisplay;
+		adjusted_mode->crtc_hblank_end = adjusted_mode->crtc_htotal;

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 824

Quoted 10 lines
drivers/gpu/drm/i915/intel_dsi.c +10 -0
+
+		adjusted_mode->crtc_vsync_start =
+					vfp + adjusted_mode->crtc_vdisplay;
+		adjusted_mode->crtc_vsync_end =
+					vsync + adjusted_mode->crtc_vsync_start;
+		adjusted_mode->crtc_vblank_start = adjusted_mode->crtc_vdisplay;
+		adjusted_mode->crtc_vblank_end = adjusted_mode->crtc_vtotal;
+
+		pipe_config->pipe_bpp =
+			mipi_dsi_pixel_format_to_bpp(intel_dsi->pixel_format);

I guess this should be read from the register.

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 837

Quoted 13 lines
drivers/gpu/drm/i915/intel_dsi.c +13 -0
+
+		dsi_pipe = BXT_PORT_TO_PIPE(I915_READ(MIPI_CTRL(port)));
+		if (dsi_pipe > PIPE_C) {
+			DRM_ERROR("Invalid PIPE configured\n");
+			break;
+		}
+
+		tmp = I915_READ(PIPESRC(dsi_pipe));
+		pipe_config->pipe_src_h = (tmp & 0xffff) + 1;
+		pipe_config->pipe_src_w = ((tmp >> 16) & 0xffff) + 1;
+
+		pipe_config->base.mode.vdisplay = pipe_config->pipe_src_h;
+		pipe_config->base.mode.hdisplay = pipe_config->pipe_src_w;

This part is already covered by intel_get_pipe_src_size() for DSI.

BR,
Jani.

Quoted 20 lines
drivers/gpu/drm/i915/intel_dsi.c +9 -0
+		break;
+	}
+}
+
+
 static void intel_dsi_get_config(struct intel_encoder *encoder,
 				 struct intel_crtc_state *pipe_config)
 {
+	struct drm_device *dev = encoder->base.dev;
 	u32 pclk;
 	DRM_DEBUG_KMS("\n");
 
 	pipe_config->has_dsi_encoder = true;
 
+	if (IS_BROXTON(dev))
+		bxt_dsi_get_pipe_config(encoder, pipe_config);
+
 	/*
 	 * DPLL_MD is not used in case of DSI, reading will get some default value
 	 * set dpll_md = 0
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#21 lore raw

Jani,

Thanks for the review comments. Addressing them in the next patch ver.

Quoted 96 lines · Jani Nikula
Quoted 83 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers.

The required changes are added for BXT in intel_dsi_get_config (encoder->get_config).

---
Previously reviewed at https://patchwork.freedesktop.org/patch/75301/

  drivers/gpu/drm/i915/i915_reg.h      |    1 +
  drivers/gpu/drm/i915/intel_display.c |   44 ++++++++++++++
  drivers/gpu/drm/i915/intel_dsi.c     |  104 ++++++++++++++++++++++++++++++++++
  3 files changed, 149 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +1 -0
index c839ce9..da3cdef 100644
@@ -8210,6 +8210,7 @@ enum skl_disp_power_wells {
  #define  BXT_PIPE_SELECT_SHIFT				7
  #define  BXT_PIPE_SELECT_MASK				(7 << 7)
  #define  BXT_PIPE_SELECT(pipe)				((pipe) << 7)
+#define  BXT_PORT_TO_PIPE(ctrl)		((ctrl & BXT_PIPE_SELECT_MASK) >> 7)
 
  #define _MIPIA_DATA_ADDRESS		(dev_priv->mipi_mmio_base + 0xb108)
  #define _MIPIC_DATA_ADDRESS		(dev_priv->mipi_mmio_base + 0xb908)
drivers/gpu/drm/i915/intel_display.c +34 -0
index 29aa64b..c0627d6 100644
@@ -9941,11 +9941,40 @@ static void haswell_get_ddi_port_state(struct intel_crtc *crtc,
  	}
  }
 
+struct intel_encoder *bxt_get_dsi_encoder_for_crtc(struct intel_crtc *crtc,
+					struct intel_crtc_state *pipe_config)
+{
+	struct drm_device *dev = crtc->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_encoder *intel_encoder;
+	struct intel_dsi *intel_dsi;
+	enum port port;
+	u32 tmp;
+
+	for_each_intel_encoder(dev, intel_encoder) {
+		if (intel_encoder->type == INTEL_OUTPUT_DSI) {
+			intel_dsi = enc_to_intel_dsi(&intel_encoder->base);
+			for_each_dsi_port(port, intel_dsi->ports) {
+				if (!(I915_READ(BXT_MIPI_PORT_CTRL(port)) &
+								DPI_ENABLE))
+					break;
+
+				tmp = I915_READ(MIPI_CTRL(port));
+				if ((tmp & BXT_PIPE_SELECT_MASK) ==
+						BXT_PIPE_SELECT(crtc->pipe))
+					return intel_encoder;
+			}
+		}
+	}
+	return NULL;
+}
+
  static bool haswell_get_pipe_config(struct intel_crtc *crtc,
  				    struct intel_crtc_state *pipe_config)
  {
  	struct drm_device *dev = crtc->base.dev;
  	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_encoder *intel_encoder, *attached_encoder = NULL;
  	enum intel_display_power_domain power_domain;
  	unsigned long power_domain_mask;
  	bool active;
@@ -9965,6 +9994,21 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc,
  		WARN_ON(active && pipe_config->has_dsi_encoder);
  		if (pipe_config->has_dsi_encoder)
  			active = true;
+
+		for_each_encoder_on_crtc(dev, &crtc->base, intel_encoder)
+			attached_encoder = intel_encoder;
+
+		/*

drivers/gpu/drm/i915/intel_display.c static bool haswell_get_pipe_config(struct intel_crtc *crtc, · line 10011

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+		 * attached_encoder will be NULL, if there is no modeset from
+		 * the kernel bootup.
+		 */
+		if (!attached_encoder && pipe_config->has_dsi_encoder)
+			attached_encoder =
+				bxt_get_dsi_encoder_for_crtc(crtc, pipe_config);
+
+		if (attached_encoder && attached_encoder->get_config)
+			attached_encoder->get_config(attached_encoder,
+								pipe_config);

No, you must not add a new call to the encoder->get_config() hook.

Quoted 7 lines

haswell_get_pipe_config() is called through the dev_priv->display.get_pipe_config() function pointer. This happens in check_crtc_state() and intel_modeset_readout_hw_state(). In both places, encoder->get_config() is called afterwards, if encoder->get_hw_state() returns true for the encoder.

The infrastructure is there, you only need to update DSI ->get_config().

Thanks for catching such a redundant lines. dropping them.

Quoted 34 lines
Quoted 34 lines
drivers/gpu/drm/i915/intel_display.c +0 -0
  	}
 
  	if (!active)
drivers/gpu/drm/i915/intel_dsi.c +22 -0
index 0de74e1..69a801e 100644
@@ -740,14 +740,118 @@ out_put_power:
  	return active;
  }
 
+/* return pixels equvalent to txbyteclkhs */
+static u16 pixels_from_txbyteclkhs(u16 clk_hs, int bpp, int lane_count,
+		       u16 burst_mode_ratio)
+{
+	return DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
+						(bpp * burst_mode_ratio));
+}
+
+static void bxt_dsi_get_pipe_config(struct intel_encoder *encoder,
+				 struct intel_crtc_state *pipe_config)
+{
+	struct drm_device *dev = encoder->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_display_mode *adjusted_mode =
+					&pipe_config->base.adjusted_mode;
+	struct intel_dsi *intel_dsi =
+				enc_to_intel_dsi(&encoder->base);
+	unsigned int bpp = mipi_dsi_pixel_format_to_bpp(intel_dsi->pixel_format);
+	unsigned int lane_count = intel_dsi->lane_count;
+	enum port port;
+	enum pipe dsi_pipe;
+	u16 hactive, hfp, hsync, hbp, vfp, vsync, vbp;
Quoted 8 lines

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 769

Quoted 5 lines
drivers/gpu/drm/i915/intel_dsi.c +5 -0
+	uint32_t tmp;
+
+	for_each_dsi_port(port, intel_dsi->ports) {
+		if (!(I915_READ(BXT_MIPI_PORT_CTRL(port)) & DPI_ENABLE))
+			continue;

Please use the loop just for figuring out the port to use for reading the registers. After that, you can do the rest at a lower indentation level, and it's clearer the rest only gets done once, for one port.

I take this suggestion for reducing the indentation level. including it next patch ver.

Quoted 49 lines
Quoted 49 lines
drivers/gpu/drm/i915/intel_dsi.c +48 -0
+
+		/* In terms of pixels */
+		adjusted_mode->crtc_hdisplay =
+					I915_READ(BXT_MIPI_TRANS_HACTIVE(port));
+		adjusted_mode->crtc_vdisplay =
+					I915_READ(BXT_MIPI_TRANS_VACTIVE(port));
+		adjusted_mode->crtc_vtotal =
+					I915_READ(BXT_MIPI_TRANS_VTOTAL(port));
+
+		hactive = adjusted_mode->crtc_hdisplay;
+		hfp = I915_READ(MIPI_HFP_COUNT(port));
+
+		/*
+		 * meaningful for video mode non-burst sync pulse mode only,
+		 * can be zero for non-burst sync events and burst modes
+		 */
+		hsync = I915_READ(MIPI_HSYNC_PADDING_COUNT(port));
+		hbp = I915_READ(MIPI_HBP_COUNT(port));
+
+		/* horizontal values are in terms of high speed byte clock */
+		hfp = pixels_from_txbyteclkhs(hfp, bpp, lane_count,
+						intel_dsi->burst_mode_ratio);
+		hsync = pixels_from_txbyteclkhs(hsync, bpp, lane_count,
+						intel_dsi->burst_mode_ratio);
+		hbp = pixels_from_txbyteclkhs(hbp, bpp, lane_count,
+						intel_dsi->burst_mode_ratio);
+
+		if (intel_dsi->dual_link) {
+			hfp *= 2;
+			hsync *= 2;
+			hbp *= 2;
+		}
+
+		/* vertical values are in terms of lines */
+		vfp = I915_READ(MIPI_VFP_COUNT(port));
+		vsync = I915_READ(MIPI_VSYNC_PADDING_COUNT(port));
+		vbp = I915_READ(MIPI_VBP_COUNT(port));
+
+		adjusted_mode->crtc_htotal = hactive + hfp + hsync + hbp;
+		adjusted_mode->crtc_hsync_start =
+					hfp + adjusted_mode->crtc_hdisplay;
+		adjusted_mode->crtc_hsync_end =
+					hsync + adjusted_mode->crtc_hsync_start;
+		adjusted_mode->crtc_hblank_start = adjusted_mode->crtc_hdisplay;
+		adjusted_mode->crtc_hblank_end = adjusted_mode->crtc_htotal;
+
+		adjusted_mode->crtc_vsync_start =
+					vfp + adjusted_mode->crtc_vdisplay;
Quoted 8 lines

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 824

Quoted 7 lines
drivers/gpu/drm/i915/intel_dsi.c +7 -0
+		adjusted_mode->crtc_vsync_end =
+					vsync + adjusted_mode->crtc_vsync_start;
+		adjusted_mode->crtc_vblank_start = adjusted_mode->crtc_vdisplay;
+		adjusted_mode->crtc_vblank_end = adjusted_mode->crtc_vtotal;
+
+		pipe_config->pipe_bpp =
+			mipi_dsi_pixel_format_to_bpp(intel_dsi->pixel_format);

I guess this should be read from the register.

Can be read from the port register. Will do it.

Quoted 15 lines
Quoted 4 lines
drivers/gpu/drm/i915/intel_dsi.c +3 -0
+
+		dsi_pipe = BXT_PORT_TO_PIPE(I915_READ(MIPI_CTRL(port)));
+		if (dsi_pipe > PIPE_C) {

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 837

Quoted 10 lines
drivers/gpu/drm/i915/intel_dsi.c +10 -0
+			DRM_ERROR("Invalid PIPE configured\n");
+			break;
+		}
+
+		tmp = I915_READ(PIPESRC(dsi_pipe));
+		pipe_config->pipe_src_h = (tmp & 0xffff) + 1;
+		pipe_config->pipe_src_w = ((tmp >> 16) & 0xffff) + 1;
+
+		pipe_config->base.mode.vdisplay = pipe_config->pipe_src_h;
+		pipe_config->base.mode.hdisplay = pipe_config->pipe_src_w;

This part is already covered by intel_get_pipe_src_size() for DSI.

Dropping those lines from get_config.

Quoted 26 lines

BR,
Jani.

Quoted 20 lines
drivers/gpu/drm/i915/intel_dsi.c +9 -0
+		break;
+	}
+}
+
+
  static void intel_dsi_get_config(struct intel_encoder *encoder,
  				 struct intel_crtc_state *pipe_config)
  {
+	struct drm_device *dev = encoder->base.dev;
  	u32 pclk;
  	DRM_DEBUG_KMS("\n");
 
  	pipe_config->has_dsi_encoder = true;
 
+	if (IS_BROXTON(dev))
+		bxt_dsi_get_pipe_config(encoder, pipe_config);
+
  	/*
  	 * DPLL_MD is not used in case of DSI, reading will get some default value
  	 * set dpll_md = 0
Signature · 7 lines
Thanks,
--Ram

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#22 lore raw

PATCH 1/2 drm/i915: Sharing the pixel_format_from_vbt to whole i915

Shared the function pixel_format_from_vbt for whole display module. Function declaration is added to intel_dsi.h.

---
 drivers/gpu/drm/i915/intel_dsi.h           |    1 +
 drivers/gpu/drm/i915/intel_dsi_panel_vbt.c |    2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dsi.h +1 -0
index ec58ead..9612916 100644
@@ -126,6 +126,7 @@ static inline struct intel_dsi *enc_to_intel_dsi(struct drm_encoder *encoder)
 	return container_of(encoder, struct intel_dsi, base.base);
 }
 
+enum mipi_dsi_pixel_format pixel_format_from_vbt(u32 fmt);
 bool intel_dsi_pll_is_enabled(struct drm_i915_private *dev_priv);
 extern void intel_enable_dsi_pll(struct intel_encoder *encoder);
 extern void intel_disable_dsi_pll(struct intel_encoder *encoder);
drivers/gpu/drm/i915/intel_dsi_panel_vbt.c +1 -1
index 8302a97..d78d59c 100644
@@ -413,7 +413,7 @@ static const struct drm_panel_funcs vbt_panel_funcs = {
 };
 
 /* XXX: This should be done when parsing the VBT in intel_bios.c */
-static enum mipi_dsi_pixel_format pixel_format_from_vbt(u32 fmt)
+enum mipi_dsi_pixel_format pixel_format_from_vbt(u32 fmt)
 {
 	/* It just so happens the VBT matches register contents. */
 	switch (fmt) {
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#23 lore raw

PATCH 2/2 drm/i915/BXT: Get pipe conf from the port registers

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers.

The required changes are added for BXT in intel_dsi_get_config (encoder->get_config).

v2: Addressed the Jani's comments

    -removed the redundant call to encoder->get_config
    -read bpp from port register
    -removed retrival of src_size from encoder->get_config

---
Previously reviewed at https://patchwork.freedesktop.org/patch/75301/

 drivers/gpu/drm/i915/intel_dsi.c |   99 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 99 insertions(+)
drivers/gpu/drm/i915/intel_dsi.c +99 -0
index 0de74e1..2117187 100644
@@ -46,6 +46,11 @@ static const struct {
 	},
 };
 
+enum mipi_dsi_pixel_format reg_to_pixel_format(u32 fmt)
+{
+	return pixel_format_from_vbt(fmt);
+}
+
 static void wait_for_dsi_fifo_empty(struct intel_dsi *intel_dsi, enum port port)
 {
 	struct drm_encoder *encoder = &intel_dsi->base.base;
@@ -740,14 +745,108 @@ out_put_power:
 	return active;
 }
 
+/* return pixels equvalent to txbyteclkhs */
+static u16 pixels_from_txbyteclkhs(u16 clk_hs, int bpp, int lane_count,
+		       u16 burst_mode_ratio)
+{
+	return DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
+						(bpp * burst_mode_ratio));
+}
+
+static void bxt_dsi_get_pipe_config(struct intel_encoder *encoder,
+				 struct intel_crtc_state *pipe_config)
+{
+	struct drm_device *dev = encoder->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_display_mode *adjusted_mode =
+					&pipe_config->base.adjusted_mode;
+	struct intel_dsi *intel_dsi = enc_to_intel_dsi(&encoder->base);
+	unsigned int lane_count = intel_dsi->lane_count;
+	unsigned int bpp, fmt;
+	enum port port;
+	u16 hactive, hfp, hsync, hbp, vfp, vsync, vbp;
+
+	/*
+	 * Atleast one port is active as encoder->get_config called only if
+	 * encoder->get_hw_state() returns true.
+	 */
+	for_each_dsi_port(port, intel_dsi->ports) {
+		if (!(I915_READ(BXT_MIPI_PORT_CTRL(port)) & DPI_ENABLE))
+			continue;
+		break;
+	}
+
+	fmt = I915_READ(MIPI_DSI_FUNC_PRG(port)) & VID_MODE_FORMAT_MASK;
+	pipe_config->pipe_bpp = reg_to_pixel_format(fmt);
+
+	bpp = mipi_dsi_pixel_format_to_bpp(pipe_config->pipe_bpp);
+
+	/* In terms of pixels */
+	adjusted_mode->crtc_hdisplay =
+				I915_READ(BXT_MIPI_TRANS_HACTIVE(port));
+	adjusted_mode->crtc_vdisplay =
+				I915_READ(BXT_MIPI_TRANS_VACTIVE(port));
+	adjusted_mode->crtc_vtotal =
+				I915_READ(BXT_MIPI_TRANS_VTOTAL(port));
+
+	hactive = adjusted_mode->crtc_hdisplay;
+	hfp = I915_READ(MIPI_HFP_COUNT(port));
+
+	/*
+	 * meaningful for video mode non-burst sync pulse mode only,
+	 * can be zero for non-burst sync events and burst modes
+	 */
+	hsync = I915_READ(MIPI_HSYNC_PADDING_COUNT(port));
+	hbp = I915_READ(MIPI_HBP_COUNT(port));
+
+	/* horizontal values are in terms of high speed byte clock */
+	hfp = pixels_from_txbyteclkhs(hfp, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+	hsync = pixels_from_txbyteclkhs(hsync, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+	hbp = pixels_from_txbyteclkhs(hbp, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+
+	if (intel_dsi->dual_link) {
+		hfp *= 2;
+		hsync *= 2;
+		hbp *= 2;
+	}
+
+	/* vertical values are in terms of lines */
+	vfp = I915_READ(MIPI_VFP_COUNT(port));
+	vsync = I915_READ(MIPI_VSYNC_PADDING_COUNT(port));
+	vbp = I915_READ(MIPI_VBP_COUNT(port));
+
+	adjusted_mode->crtc_htotal = hactive + hfp + hsync + hbp;
+	adjusted_mode->crtc_hsync_start =
+				hfp + adjusted_mode->crtc_hdisplay;
+	adjusted_mode->crtc_hsync_end =
+				hsync + adjusted_mode->crtc_hsync_start;
+	adjusted_mode->crtc_hblank_start = adjusted_mode->crtc_hdisplay;
+	adjusted_mode->crtc_hblank_end = adjusted_mode->crtc_htotal;
+
+	adjusted_mode->crtc_vsync_start =
+				vfp + adjusted_mode->crtc_vdisplay;
+	adjusted_mode->crtc_vsync_end =
+				vsync + adjusted_mode->crtc_vsync_start;
+	adjusted_mode->crtc_vblank_start = adjusted_mode->crtc_vdisplay;
+	adjusted_mode->crtc_vblank_end = adjusted_mode->crtc_vtotal;
+}
+
+
 static void intel_dsi_get_config(struct intel_encoder *encoder,
 				 struct intel_crtc_state *pipe_config)
 {
+	struct drm_device *dev = encoder->base.dev;
 	u32 pclk;
 	DRM_DEBUG_KMS("\n");
 
 	pipe_config->has_dsi_encoder = true;
 
+	if (IS_BROXTON(dev))
+		bxt_dsi_get_pipe_config(encoder, pipe_config);
+
 	/*
 	 * DPLL_MD is not used in case of DSI, reading will get some default value
 	 * set dpll_md = 0
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#24 lore raw

Jani/Daniel,

I am working on implementing the pipe_config compare as suggested by daniel at https://lists.freedesktop.org/archives/intel-gfx/2016-March/091148.html

But I think this patch need not wait for that change. Either way this patch is required. We can continue review on this and proceed further.

Quoted 145 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers.

The required changes are added for BXT in intel_dsi_get_config (encoder->get_config).

v2: Addressed the Jani's comments

     -removed the redundant call to encoder->get_config
     -read bpp from port register
     -removed retrival of src_size from encoder->get_config

---
Previously reviewed at https://patchwork.freedesktop.org/patch/75301/

  drivers/gpu/drm/i915/intel_dsi.c |   99 ++++++++++++++++++++++++++++++++++++++
  1 file changed, 99 insertions(+)
drivers/gpu/drm/i915/intel_dsi.c +99 -0
index 0de74e1..2117187 100644
@@ -46,6 +46,11 @@ static const struct {
  	},
  };
 
+enum mipi_dsi_pixel_format reg_to_pixel_format(u32 fmt)
+{
+	return pixel_format_from_vbt(fmt);
+}
+
  static void wait_for_dsi_fifo_empty(struct intel_dsi *intel_dsi, enum port port)
  {
  	struct drm_encoder *encoder = &intel_dsi->base.base;
@@ -740,14 +745,108 @@ out_put_power:
  	return active;
  }
 
+/* return pixels equvalent to txbyteclkhs */
+static u16 pixels_from_txbyteclkhs(u16 clk_hs, int bpp, int lane_count,
+		       u16 burst_mode_ratio)
+{
+	return DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
+						(bpp * burst_mode_ratio));
+}
+
+static void bxt_dsi_get_pipe_config(struct intel_encoder *encoder,
+				 struct intel_crtc_state *pipe_config)
+{
+	struct drm_device *dev = encoder->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_display_mode *adjusted_mode =
+					&pipe_config->base.adjusted_mode;
+	struct intel_dsi *intel_dsi = enc_to_intel_dsi(&encoder->base);
+	unsigned int lane_count = intel_dsi->lane_count;
+	unsigned int bpp, fmt;
+	enum port port;
+	u16 hactive, hfp, hsync, hbp, vfp, vsync, vbp;
+
+	/*
+	 * Atleast one port is active as encoder->get_config called only if
+	 * encoder->get_hw_state() returns true.
+	 */
+	for_each_dsi_port(port, intel_dsi->ports) {
+		if (!(I915_READ(BXT_MIPI_PORT_CTRL(port)) & DPI_ENABLE))
+			continue;
+		break;
+	}
+
+	fmt = I915_READ(MIPI_DSI_FUNC_PRG(port)) & VID_MODE_FORMAT_MASK;
+	pipe_config->pipe_bpp = reg_to_pixel_format(fmt);
+
+	bpp = mipi_dsi_pixel_format_to_bpp(pipe_config->pipe_bpp);
+
+	/* In terms of pixels */
+	adjusted_mode->crtc_hdisplay =
+				I915_READ(BXT_MIPI_TRANS_HACTIVE(port));
+	adjusted_mode->crtc_vdisplay =
+				I915_READ(BXT_MIPI_TRANS_VACTIVE(port));
+	adjusted_mode->crtc_vtotal =
+				I915_READ(BXT_MIPI_TRANS_VTOTAL(port));
+
+	hactive = adjusted_mode->crtc_hdisplay;
+	hfp = I915_READ(MIPI_HFP_COUNT(port));
+
+	/*
+	 * meaningful for video mode non-burst sync pulse mode only,
+	 * can be zero for non-burst sync events and burst modes
+	 */
+	hsync = I915_READ(MIPI_HSYNC_PADDING_COUNT(port));
+	hbp = I915_READ(MIPI_HBP_COUNT(port));
+
+	/* horizontal values are in terms of high speed byte clock */
+	hfp = pixels_from_txbyteclkhs(hfp, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+	hsync = pixels_from_txbyteclkhs(hsync, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+	hbp = pixels_from_txbyteclkhs(hbp, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+
+	if (intel_dsi->dual_link) {
+		hfp *= 2;
+		hsync *= 2;
+		hbp *= 2;
+	}
+
+	/* vertical values are in terms of lines */
+	vfp = I915_READ(MIPI_VFP_COUNT(port));
+	vsync = I915_READ(MIPI_VSYNC_PADDING_COUNT(port));
+	vbp = I915_READ(MIPI_VBP_COUNT(port));
+
+	adjusted_mode->crtc_htotal = hactive + hfp + hsync + hbp;
+	adjusted_mode->crtc_hsync_start =
+				hfp + adjusted_mode->crtc_hdisplay;
+	adjusted_mode->crtc_hsync_end =
+				hsync + adjusted_mode->crtc_hsync_start;
+	adjusted_mode->crtc_hblank_start = adjusted_mode->crtc_hdisplay;
+	adjusted_mode->crtc_hblank_end = adjusted_mode->crtc_htotal;
+
+	adjusted_mode->crtc_vsync_start =
+				vfp + adjusted_mode->crtc_vdisplay;
+	adjusted_mode->crtc_vsync_end =
+				vsync + adjusted_mode->crtc_vsync_start;
+	adjusted_mode->crtc_vblank_start = adjusted_mode->crtc_vdisplay;
+	adjusted_mode->crtc_vblank_end = adjusted_mode->crtc_vtotal;
+}
+
+
  static void intel_dsi_get_config(struct intel_encoder *encoder,
  				 struct intel_crtc_state *pipe_config)
  {
+	struct drm_device *dev = encoder->base.dev;
  	u32 pclk;
  	DRM_DEBUG_KMS("\n");
 
  	pipe_config->has_dsi_encoder = true;
 
+	if (IS_BROXTON(dev))
+		bxt_dsi_get_pipe_config(encoder, pipe_config);
+
  	/*
  	 * DPLL_MD is not used in case of DSI, reading will get some default value
  	 * set dpll_md = 0
Signature · 7 lines
Thanks,
--Ram

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#25 lore raw
In reply to Ramalingam C
Quoted 25 lines · Ramalingam C

At BXT DSI, PIPE registers are inactive. So we can't get the PIPE's mode parameters from them. The possible option is retriving them from the PORT registers.

The required changes are added for BXT in intel_dsi_get_config (encoder->get_config).

v2: Addressed the Jani's comments

    -removed the redundant call to encoder->get_config
    -read bpp from port register
    -removed retrival of src_size from encoder->get_config

---
Previously reviewed at https://patchwork.freedesktop.org/patch/75301/

 drivers/gpu/drm/i915/intel_dsi.c |   99 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 99 insertions(+)
drivers/gpu/drm/i915/intel_dsi.c +0 -0
index 0de74e1..2117187 100644

drivers/gpu/drm/i915/intel_dsi.c static const struct { · line 53

Quoted 8 lines
drivers/gpu/drm/i915/intel_dsi.c +5 -0
@@ -46,6 +46,11 @@ static const struct {
 	},
 };
 
+enum mipi_dsi_pixel_format reg_to_pixel_format(u32 fmt)
+{
+	return pixel_format_from_vbt(fmt);
+}
+

See reply to preceding patch.

Quoted 26 lines
drivers/gpu/drm/i915/intel_dsi.c +19 -0
 static void wait_for_dsi_fifo_empty(struct intel_dsi *intel_dsi, enum port port)
 {
 	struct drm_encoder *encoder = &intel_dsi->base.base;
@@ -740,14 +745,108 @@ out_put_power:
 	return active;
 }
 
+/* return pixels equvalent to txbyteclkhs */
+static u16 pixels_from_txbyteclkhs(u16 clk_hs, int bpp, int lane_count,
+		       u16 burst_mode_ratio)
+{
+	return DIV_ROUND_UP((clk_hs * lane_count * 8 * 100),
+						(bpp * burst_mode_ratio));
+}
+
+static void bxt_dsi_get_pipe_config(struct intel_encoder *encoder,
+				 struct intel_crtc_state *pipe_config)
+{
+	struct drm_device *dev = encoder->base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_display_mode *adjusted_mode =
+					&pipe_config->base.adjusted_mode;
+	struct intel_dsi *intel_dsi = enc_to_intel_dsi(&encoder->base);
+	unsigned int lane_count = intel_dsi->lane_count;
+	unsigned int bpp, fmt;
+	enum port port;

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 776

Quoted 10 lines
drivers/gpu/drm/i915/intel_dsi.c +10 -0
+	u16 hactive, hfp, hsync, hbp, vfp, vsync, vbp;
+
+	/*
+	 * Atleast one port is active as encoder->get_config called only if
+	 * encoder->get_hw_state() returns true.
+	 */
+	for_each_dsi_port(port, intel_dsi->ports) {
+		if (!(I915_READ(BXT_MIPI_PORT_CTRL(port)) & DPI_ENABLE))
+			continue;
+		break;

In other words,

	if (I915_READ(BXT_MIPI_PORT_CTRL(port)) & DPI_ENABLE)
		break;
Quoted 48 lines
drivers/gpu/drm/i915/intel_dsi.c +48 -0
+	}
+
+	fmt = I915_READ(MIPI_DSI_FUNC_PRG(port)) & VID_MODE_FORMAT_MASK;
+	pipe_config->pipe_bpp = reg_to_pixel_format(fmt);
+
+	bpp = mipi_dsi_pixel_format_to_bpp(pipe_config->pipe_bpp);
+
+	/* In terms of pixels */
+	adjusted_mode->crtc_hdisplay =
+				I915_READ(BXT_MIPI_TRANS_HACTIVE(port));
+	adjusted_mode->crtc_vdisplay =
+				I915_READ(BXT_MIPI_TRANS_VACTIVE(port));
+	adjusted_mode->crtc_vtotal =
+				I915_READ(BXT_MIPI_TRANS_VTOTAL(port));
+
+	hactive = adjusted_mode->crtc_hdisplay;
+	hfp = I915_READ(MIPI_HFP_COUNT(port));
+
+	/*
+	 * meaningful for video mode non-burst sync pulse mode only,
+	 * can be zero for non-burst sync events and burst modes
+	 */
+	hsync = I915_READ(MIPI_HSYNC_PADDING_COUNT(port));
+	hbp = I915_READ(MIPI_HBP_COUNT(port));
+
+	/* horizontal values are in terms of high speed byte clock */
+	hfp = pixels_from_txbyteclkhs(hfp, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+	hsync = pixels_from_txbyteclkhs(hsync, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+	hbp = pixels_from_txbyteclkhs(hbp, bpp, lane_count,
+					intel_dsi->burst_mode_ratio);
+
+	if (intel_dsi->dual_link) {
+		hfp *= 2;
+		hsync *= 2;
+		hbp *= 2;
+	}
+
+	/* vertical values are in terms of lines */
+	vfp = I915_READ(MIPI_VFP_COUNT(port));
+	vsync = I915_READ(MIPI_VSYNC_PADDING_COUNT(port));
+	vbp = I915_READ(MIPI_VBP_COUNT(port));
+
+	adjusted_mode->crtc_htotal = hactive + hfp + hsync + hbp;
+	adjusted_mode->crtc_hsync_start =
+				hfp + adjusted_mode->crtc_hdisplay;
+	adjusted_mode->crtc_hsync_end =

drivers/gpu/drm/i915/intel_dsi.c out_put_power: · line 834

Quoted 10 lines
drivers/gpu/drm/i915/intel_dsi.c +10 -0
+				hsync + adjusted_mode->crtc_hsync_start;
+	adjusted_mode->crtc_hblank_start = adjusted_mode->crtc_hdisplay;
+	adjusted_mode->crtc_hblank_end = adjusted_mode->crtc_htotal;
+
+	adjusted_mode->crtc_vsync_start =
+				vfp + adjusted_mode->crtc_vdisplay;
+	adjusted_mode->crtc_vsync_end =
+				vsync + adjusted_mode->crtc_vsync_start;
+	adjusted_mode->crtc_vblank_start = adjusted_mode->crtc_vdisplay;
+	adjusted_mode->crtc_vblank_end = adjusted_mode->crtc_vtotal;

As I suggested on IRC, to make progress please keep everything that we can calculate accurately (and hence get no warnings about the diff for them) and leave out all that is not accurate.

We need to get back to them too, but currently there seems to be disagreement on what is the best/feasible/possible option.

BR,
Jani.

Quoted 18 lines
drivers/gpu/drm/i915/intel_dsi.c +7 -0
+}
+
+
 static void intel_dsi_get_config(struct intel_encoder *encoder,
 				 struct intel_crtc_state *pipe_config)
 {
+	struct drm_device *dev = encoder->base.dev;
 	u32 pclk;
 	DRM_DEBUG_KMS("\n");
 
 	pipe_config->has_dsi_encoder = true;
 
+	if (IS_BROXTON(dev))
+		bxt_dsi_get_pipe_config(encoder, pipe_config);
+
 	/*
 	 * DPLL_MD is not used in case of DSI, reading will get some default value
 	 * set dpll_md = 0
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#26 lore raw

Re: [PATCH 1/2] drm/i915: Sharing the pixel_format_from_vbt to whole i915

In reply to Ramalingam C
Quoted 26 lines · Ramalingam C

Shared the function pixel_format_from_vbt for whole display module. Function declaration is added to intel_dsi.h.

---
 drivers/gpu/drm/i915/intel_dsi.h           |    1 +
 drivers/gpu/drm/i915/intel_dsi_panel_vbt.c |    2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dsi.h +1 -0
index ec58ead..9612916 100644
@@ -126,6 +126,7 @@ static inline struct intel_dsi *enc_to_intel_dsi(struct drm_encoder *encoder)
 	return container_of(encoder, struct intel_dsi, base.base);
 }
 
+enum mipi_dsi_pixel_format pixel_format_from_vbt(u32 fmt);
 bool intel_dsi_pll_is_enabled(struct drm_i915_private *dev_priv);
 extern void intel_enable_dsi_pll(struct intel_encoder *encoder);
 extern void intel_disable_dsi_pll(struct intel_encoder *encoder);
drivers/gpu/drm/i915/intel_dsi_panel_vbt.c +0 -0
index 8302a97..d78d59c 100644

drivers/gpu/drm/i915/intel_dsi_panel_vbt.c static const struct drm_panel_funcs vbt_panel_funcs = { · line 416

Quoted 5 lines
drivers/gpu/drm/i915/intel_dsi_panel_vbt.c +1 -1
@@ -413,7 +413,7 @@ static const struct drm_panel_funcs vbt_panel_funcs = {
 };
 
 /* XXX: This should be done when parsing the VBT in intel_bios.c */
-static enum mipi_dsi_pixel_format pixel_format_from_vbt(u32 fmt)
+enum mipi_dsi_pixel_format pixel_format_from_vbt(u32 fmt)

I think this one ends up being nicer if you move the whole function to intel_dsi.c and name it according to being a function to convert the pixel format from the *register* data, not vbt.

BR,
Jani.

Quoted 3 lines
drivers/gpu/drm/i915/intel_dsi_panel_vbt.c +0 -0
 {
 	/* It just so happens the VBT matches register contents. */
 	switch (fmt) {
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#27 lore raw

✗ Fi.CI.BAT: failure for series starting with [2/2] drm/i915/BXT: Get pipe conf from the port registers (rev3)

In reply to Ramalingam C
== Series Details ==

== Series Details ==

Series: series starting with [2/2] drm/i915/BXT: Get pipe conf from the port registers (rev3)
URL : https://patchwork.freedesktop.org/series/5006/
State : failure

== Summary ==

  CC [M]  drivers/net/ethernet/intel/igb/e1000_mac.o
  CC [M]  drivers/net/ethernet/intel/e1000e/netdev.o
  CC [M]  drivers/net/ethernet/intel/igb/e1000_nvm.o
  CC [M]  drivers/net/ethernet/intel/e1000e/ptp.o
  CC [M]  drivers/net/ethernet/intel/igb/e1000_phy.o
  CC [M]  drivers/net/ethernet/intel/igb/e1000_mbx.o
  CC [M]  drivers/net/ethernet/intel/igb/e1000_i210.o
  CC [M]  drivers/net/ethernet/intel/igb/igb_ptp.o
  CC [M]  drivers/net/ethernet/intel/igb/igb_hwmon.o
  LD      drivers/usb/host/xhci-hcd.o
  LD      drivers/usb/host/built-in.o
  LD      drivers/usb/built-in.o
  LD [M]  drivers/net/ethernet/intel/e1000/e1000.o
  LD [M]  drivers/net/ethernet/intel/igbvf/igbvf.o

cc1: some warnings being treated as errors
scripts/Makefile.build:291: recipe for target 'drivers/gpu/drm/i915/intel_display.o' failed make[4]: *** [drivers/gpu/drm/i915/intel_display.o] Error 1 scripts/Makefile.build:440: recipe for target 'drivers/gpu/drm/i915' failed make[3]: *** [drivers/gpu/drm/i915] Error 2
scripts/Makefile.build:440: recipe for target 'drivers/gpu/drm' failed make[2]: *** [drivers/gpu/drm] Error 2
scripts/Makefile.build:440: recipe for target 'drivers/gpu' failed
make[1]: *** [drivers/gpu] Error 2
make[1]: *** Waiting for unfinished jobs....

  LD [M]  drivers/net/ethernet/intel/igb/igb.o
  LD [M]  drivers/net/ethernet/intel/e1000e/e1000e.o
  LD      drivers/net/ethernet/built-in.o
  LD      drivers/net/built-in.o

Makefile:962: recipe for target 'drivers' failed
make: *** [drivers] Error 2

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx