drm/i915: Fix up the CCS code
Subsystems (3)
cover drm/i915: Fix up the CCS code
From: Ville Syrjälä <[email protected]>
Looks like we ended up with a stale version of my CCS code in dinq. This series contains the remainder in smaller chunks. I also ended up adding a bunch of extra cleanup etc. on top.
The most important thing we need to get in is the change to the fb->offsets[] interpretation since that's ABI territory. The hash mode apparently doesn't require the nasty virtual address alignment tricks so that shouldn't have any ABI issues after all.
Entire series available here:git://github.com/vsyrjala/linux.git ccs_fixes
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offset
drm/i915: Skip fence alignemnt check for the CCS plane drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCS
drm/i915: Add a comment exlaining CCS hsub/vsub
drm/i915: Nuke a pointless unreachable()
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites drm/i915: Clean up the sprite modifier checks
drm/i915: Add CCS capability for sprites
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
drm: Fix modifiers_property kernel doc
drm: Check that the plane supports the request format+modifier combo drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()
drivers/gpu/drm/drm_atomic.c | 8 +- drivers/gpu/drm/drm_crtc.c | 8 +- drivers/gpu/drm/drm_crtc_internal.h | 4 +- drivers/gpu/drm/drm_plane.c | 31 +++++-- drivers/gpu/drm/i915/i915_reg.h | 8 +- drivers/gpu/drm/i915/intel_display.c | 145 +++++++++++++++++---------------- drivers/gpu/drm/i915/intel_drv.h | 2 + drivers/gpu/drm/i915/intel_engine_cs.c | 13 +++ drivers/gpu/drm/i915/intel_pm.c | 27 +++--- drivers/gpu/drm/i915/intel_sprite.c | 102 +++++++++++++++-------- include/drm/drm_mode_config.h | 2 +- 11 files changed, 217 insertions(+), 133 deletions(-)
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 1/12 drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear offset
From: Ville Syrjälä <[email protected]>
Userspace wants to treat fb->offsets[] as raw byte offsets into the gem bo. Adjust the kernel code to match.
--- drivers/gpu/drm/i915/intel_display.c | 108 +++++++++++++++++++++-------------- 1 file changed, 64 insertions(+), 44 deletions(-)
drivers/gpu/drm/i915/intel_display.c +64 -44
@@ -2288,17 +2288,13 @@ void intel_add_fb_offsets(int *x, int *y, } } -/* -* Input tile dimensions and pitch must already be -* rotated to match x and y, and in pixel units. -*/ -static u32 _intel_adjust_tile_offset (int *x,int *y, -unsigned int tile_width, -unsigned int tile_height, -unsigned int tile_size, -unsigned int pitch_tiles, -u32 old_offset, -u32 new_offset) +static u32 __intel_adjust_tile_offset (int *x,int *y, +unsigned int tile_width, +unsigned int tile_height, +unsigned int tile_size, +unsigned int pitch_tiles, +u32 old_offset, +u32 new_offset) {unsigned int pitch_pixels = pitch_tiles * tile_width;unsigned int tiles;
@@ -2319,18 +2315,13 @@ static u32 _intel_adjust_tile_offset(int *x, int *y,return new_offset; } -/* -* Adjust the tile offset by moving the difference into -* the x/y offsets. -*/ -static u32 intel_adjust_tile_offset (int *x,int *y, -const struct intel_plane_state *state,int plane, -u32 old_offset,u32 new_offset) +static u32 _intel_adjust_tile_offset (int *x,int *y, +const struct drm_framebuffer *fb,int plane, +unsigned int rotation, +u32 old_offset,u32 new_offset) { -const struct drm_i915_private *dev_priv =to_i915 (state->base.plane->dev); -const struct drm_framebuffer *fb = state->base.fb; +const struct drm_i915_private *dev_priv =to_i915 (fb->dev);unsigned int cpp = fb->format->cpp[plane]; -unsigned int rotation = state->base.rotation;unsigned int pitch =intel_fb_pitch (fb, plane, rotation);WARN_ON (new_offset > old_offset);
@@ -2349,9 +2340,9 @@ static u32 intel_adjust_tile_offset(int *x, int *y, pitch_tiles = pitch / (tile_width * cpp); } -_intel_adjust_tile_offset (x, y, tile_width, tile_height, - tile_size, pitch_tiles, - old_offset, new_offset); +__intel_adjust_tile_offset (x, y, tile_width, tile_height, + tile_size, pitch_tiles, + old_offset, new_offset); }else { old_offset += *y * pitch + *x * cpp;
@@ -2363,6 +2354,19 @@ static u32 intel_adjust_tile_offset(int *x, int *y, }/* +* Adjust the tile offset by moving the difference into +* the x/y offsets. +*/ +static u32 intel_adjust_tile_offset (int *x,int *y, +const struct intel_plane_state *state,int plane, +u32 old_offset,u32 new_offset) +{ +return _intel_adjust_tile_offset (x, y, state->base.fb, plane, + state->base.rotation, + old_offset, new_offset); +} + +/* * Computes the linear offset to the base tile and adjusts * x, y. bytes per pixel is assumed to be a power-of-two. *
@@ -2413,9 +2417,9 @@ static u32 _intel_compute_tile_offset(const struct drm_i915_private *dev_priv, offset = (tile_rows * pitch_tiles + tiles) * tile_size; offset_aligned = offset & ~alignment; -_intel_adjust_tile_offset (x, y, tile_width, tile_height, - tile_size, pitch_tiles, - offset, offset_aligned); +__intel_adjust_tile_offset (x, y, tile_width, tile_height, + tile_size, pitch_tiles, + offset, offset_aligned); }else { offset = *y * pitch + *x * cpp; offset_aligned = offset & ~alignment;
@@ -2447,16 +2451,24 @@ u32 intel_compute_tile_offset(int *x, int *y, rotation, alignment); } -/* Convert the fb->offset[] linear offset into x/y offsets */ -static void intel_fb_offset_to_xy (int *x,int *y, -const struct drm_framebuffer *fb,int plane) +/* Convert the fb->offset[] into x/y offsets */ +static int intel_fb_offset_to_xy (int *x,int *y, +const struct drm_framebuffer *fb,int plane) { -unsigned int cpp = fb->format->cpp[plane]; -unsigned int pitch = fb->pitches[plane]; -u32 linear_offset = fb->offsets[plane]; +struct drm_i915_private *dev_priv =to_i915 (fb->dev); + +if (fb->modifier !=DRM_FORMAT_MOD_LINEAR && + fb->offsets[plane] %intel_tile_size (dev_priv)) +return -EINVAL ; - *y = linear_offset / pitch; - *x = linear_offset % pitch / cpp; + *x =0 ; + *y =0 ; + +_intel_adjust_tile_offset (x, y, + fb, plane,DRM_MODE_ROTATE_0 , + fb->offsets[plane],0 ); + +return 0 ; }static unsigned int intel_fb_modifier_to_tiling (uint64_t fb_modifier)
@@ -2523,12 +2535,18 @@ intel_fill_fb_info(struct drm_i915_private *dev_priv,unsigned int cpp, size;u32 offset;int x, y; +int ret; cpp = fb->format->cpp[i]; width =drm_framebuffer_plane_width (fb->width, fb, i); height =drm_framebuffer_plane_height (fb->height, fb, i); -intel_fb_offset_to_xy (&x, &y, fb, i); + ret =intel_fb_offset_to_xy (&x, &y, fb, i); +if (ret) { +DRM_DEBUG_KMS ("bad fb plane %d offset: 0x%x\n" , + i, fb->offsets[i]); +return ret; + }if ((fb->modifier ==I915_FORMAT_MOD_Y_TILED_CCS || fb->modifier == I915_FORMAT_MOD_Yf_TILED_CCS) && i ==1 ) {
@@ -2539,11 +2557,13 @@ intel_fill_fb_info(struct drm_i915_private *dev_priv,int ccs_x, ccs_y;intel_tile_dims (fb, i, &tile_width, &tile_height); + tile_width *= hsub; + tile_height *= vsub; - ccs_x = (x * hsub) % (tile_width * hsub); - ccs_y = (y * vsub) % (tile_height * vsub); - main_x = intel_fb->normal[0 ].x % (tile_width * hsub); - main_y = intel_fb->normal[0 ].y % (tile_height * vsub); + ccs_x = (x * hsub) % tile_width; + ccs_y = (y * vsub) % tile_height; + main_x = intel_fb->normal[0 ].x % tile_width; + main_y = intel_fb->normal[0 ].y % tile_height;/* * CCS doesn't have its own x/y offset register, so the intra CCS tile
@@ -2632,10 +2652,10 @@ intel_fill_fb_info(struct drm_i915_private *dev_priv,* We only keep the x/y offsets, so push all of the * gtt offset into the x/y offsets. */ -_intel_adjust_tile_offset (&x, &y, - tile_width, tile_height, - tile_size, pitch_tiles, - gtt_offset_rotated * tile_size,0 ); +__intel_adjust_tile_offset (&x, &y, + tile_width, tile_height, + tile_size, pitch_tiles, + gtt_offset_rotated * tile_size,0 ); gtt_offset_rotated += rot_info->plane[i].width * rot_info->plane[i].height;
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 2/12 drm/i915: Skip fence alignemnt check for the CCS plane
In reply to [email protected]From: Ville Syrjälä <[email protected]>
The CCS won't have the same stride as the main surface anyway so trying to guard against the fence stride not matching the CCS stride is not sensible. Just skip the fence vs. fb alignment check for the aux plane.
--- drivers/gpu/drm/i915/intel_display.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_display.c +1 -1
@@ -2589,7 +2589,7 @@ intel_fill_fb_info(struct drm_i915_private *dev_priv,* fb layout agrees with the fence layout. We already check that the * fb stride matches the fence stride elsewhere. */ -if (i915_gem_object_is_tiled (intel_fb->obj) && +if (i ==0 &&i915_gem_object_is_tiled (intel_fb->obj) && (x + width) * cpp > fb->pitches[i]) {DRM_DEBUG_KMS ("bad fb plane %d offset: 0x%x\n" , i, fb->offsets[i]);
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 02/12] drm/i915: Skip fence alignemnt check for the CCS plane
Quoted 20 lines · Ville Syrjälä
From: Ville Syrjälä <[email protected]>
The CCS won't have the same stride as the main surface anyway so trying to guard against the fence stride not matching the CCS stride is not sensible. Just skip the fence vs. fb alignment check for the aux plane.
Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>---
drivers/gpu/drm/i915/intel_display.c| 2 +-1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/i915/intel_display.c +0
-0
drivers/gpu/drm/i915/intel_display.c intel_fill_fb_info(struct drm_i915_private *dev_priv, · line 2595
Quoted 8 lines
drivers/gpu/drm/i915/intel_display.c +1
-1@@ -2589,7 +2589,7 @@ intel_fill_fb_info(struct drm_i915_private *dev_priv,* fb layout agrees with the fence layout. We already check that the * fb stride matches the fence stride elsewhere. */ -if (i915_gem_object_is_tiled (intel_fb->obj) && +if (i ==0 &&i915_gem_object_is_tiled (intel_fb->obj) && (x + width) * cpp > fb->pitches[i]) {DRM_DEBUG_KMS ("bad fb plane %d offset: 0x%x\n" , i, fb->offsets[i]);
Patch 1 is too much churn for my eyes to review, but it's
This one is
Signature · 5 lines
Ben Widawsky, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 3/12 drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for CCS
In reply to [email protected]From: Ville Syrjälä <[email protected]>
Use the LLC/eLLC hotspot avoidance mode for CCS on LLC machines. This is reported to give better performance.
Testing has indicated that we don't need to enforce any massive 2 or 4 MiB alignment for all compressed resources even though there are still plenty of stale comments in the spec suggesting that we do.
We do need to make sure every hardware unit that deals with the compressed data uses the same hash mode.
--- drivers/gpu/drm/i915/i915_reg.h | 8 +++++++- drivers/gpu/drm/i915/intel_engine_cs.c | 13 +++++++++++++ drivers/gpu/drm/i915/intel_pm.c | 27 +++++++++++++-------------- 3 files changed, 33 insertions(+), 15 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +7 -1
@@ -6909,7 +6909,7 @@ enum {# define CHICKEN3_DGMG_DONE_FIX_DISABLE (1 << 2)#define CHICKEN_PAR1_1 _MMIO(0x42080) -#define SKL_RC_HASH_OUTSIDE (1 << 15) +#define SKL_DE_COMPRESSED_HASH_MODE (1 << 15)#define DPA_MASK_VBLANK_SRD (1 << 15)#define FORCE_ARB_IDLE_PLANES (1 << 14)#define SKL_EDP_PSR_FIX_RDWRAP (1 << 3)
@@ -6982,6 +6982,7 @@ enum {# define GEN7_CSC1_RHWO_OPT_DISABLE_IN_RCC ((1<<10) | (1<<26))# define GEN9_RHWO_OPTIMIZATION_DISABLE (1<<14)#define COMMON_SLICE_CHICKEN2 _MMIO(0x7014) +# define GEN9_PBE_COMPRESSED_HASH_SELECTION (1<<13)# define GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE (1<<12)# define GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION (1<<8)# define GEN8_CSC2_SBE_VUE_CACHE_CONSERVATIVE (1<<0)
@@ -8071,6 +8072,7 @@ enum {#define GEN8_SAMPLER_POWER_BYPASS_DIS (1<<1)#define GEN9_HALF_SLICE_CHICKEN7 _MMIO(0xe194) +#define GEN9_SAMPLER_HASH_COMPRESSED_READ_ADDR (1<<8)#define GEN9_ENABLE_YV12_BUGFIX (1<<4)#define GEN9_ENABLE_GPGPU_PREEMPTION (1<<2)
@@ -9371,4 +9373,8 @@ enum skl_power_gate {#define GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_SKL 0x67F1427F/* " " */ #define GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_BXT 0x5FF101FF/* " " */ +#define MMCD_MISC_CTRL _MMIO(0x4ddc)/* skl+ */ +#define MMCD_PCLA (1 << 31) +#define MMCD_HOTSPOT_EN (1 << 27) +#endif /* _I915_REG_H_ */
drivers/gpu/drm/i915/intel_engine_cs.c +13 -0
@@ -812,6 +812,19 @@ static int gen9_init_workarounds(struct intel_engine_cs *engine)I915_WRITE (GAM_ECOCHK ,I915_READ (GAM_ECOCHK ) |ECOCHK_DIS_TLB ); +if (HAS_LLC (dev_priv)) { +/* WaCompressedResourceSamplerPbeMediaNewHashMode:skl,kbl +* +* Must match Display Engine. See +* WaCompressedResourceDisplayNewHashMode. +*/ +WA_SET_BIT_MASKED (COMMON_SLICE_CHICKEN2 , +GEN9_PBE_COMPRESSED_HASH_SELECTION ); +WA_SET_BIT_MASKED (GEN9_HALF_SLICE_CHICKEN7 , +GEN9_SAMPLER_HASH_COMPRESSED_READ_ADDR ); +WA_SET_BIT (MMCD_MISC_CTRL ,MMCD_PCLA |MMCD_HOTSPOT_EN ); + } +/* WaClearFlowControlGpgpuContextSave:skl,bxt,kbl,glk,cfl */ /* WaDisablePartialInstShootdown:skl,bxt,kbl,glk,cfl */ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
drivers/gpu/drm/i915/intel_pm.c +13 -14
@@ -58,24 +58,23 @@static void gen9_init_clock_gating (struct drm_i915_private *dev_priv) { +if (HAS_LLC (dev_priv)) { +/* +* WaCompressedResourceDisplayNewHashMode:skl,kbl +* Display WA#0390: skl,kbl +* +* Must match Sampler, Pixel Back End, and Media. See +* WaCompressedResourceSamplerPbeMediaNewHashMode. +*/ +I915_WRITE (CHICKEN_PAR1_1 , +I915_READ (CHICKEN_PAR1_1 ) | +SKL_DE_COMPRESSED_HASH_MODE ); + } +/* See Bspec note for PSR2_CTL bit 31, Wa#828:skl,bxt,kbl,cfl */ I915_WRITE (CHICKEN_PAR1_1 ,I915_READ (CHICKEN_PAR1_1 ) |SKL_EDP_PSR_FIX_RDWRAP ); -/* -* Display WA#0390: skl,bxt,kbl,glk -* -* Must match Sampler, Pixel Back End, and Media -* (0xE194 bit 8, 0x7014 bit 13, 0x4DDC bits 27 and 31). -* -* Including bits outside the page in the hash would -* require 2 (or 4?) MiB alignment of resources. Just -* assume the defaul hashing mode which only uses bits -* within the page. -*/ -I915_WRITE (CHICKEN_PAR1_1 , -I915_READ (CHICKEN_PAR1_1 ) & ~SKL_RC_HASH_OUTSIDE ); -I915_WRITE (GEN8_CONFIG0 ,I915_READ (GEN8_CONFIG0 ) |GEN9_DEFAULT_FIXES );
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 03/12] drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for CCS
Quoted 5 lines · Ville Syrjälä
From: Ville Syrjälä <[email protected]>
Use the LLC/eLLC hotspot avoidance mode for CCS on LLC machines. This is reported to give better performance.
Not seeing in the diff how this only hits eLLC machines. Am I misreading when this is needed.
Quoted 124 lines
Testing has indicated that we don't need to enforce any massive 2 or 4 MiB alignment for all compressed resources even though there are still plenty of stale comments in the spec suggesting that we do.
We do need to make sure every hardware unit that deals with the compressed data uses the same hash mode.
Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>---
drivers/gpu/drm/i915/i915_reg.h | 8 +++++++- drivers/gpu/drm/i915/intel_engine_cs.c | 13 +++++++++++++ drivers/gpu/drm/i915/intel_pm.c | 27 +++++++++++++-------------- 3 files changed, 33 insertions(+), 15 deletions(-)drivers/gpu/drm/i915/i915_reg.h +7
-1@@ -6909,7 +6909,7 @@ enum {# define CHICKEN3_DGMG_DONE_FIX_DISABLE (1 << 2)#define CHICKEN_PAR1_1 _MMIO(0x42080) -#define SKL_RC_HASH_OUTSIDE (1 << 15) +#define SKL_DE_COMPRESSED_HASH_MODE (1 << 15)#define DPA_MASK_VBLANK_SRD (1 << 15)#define FORCE_ARB_IDLE_PLANES (1 << 14)#define SKL_EDP_PSR_FIX_RDWRAP (1 << 3)@@ -6982,6 +6982,7 @@ enum {# define GEN7_CSC1_RHWO_OPT_DISABLE_IN_RCC ((1<<10) | (1<<26))# define GEN9_RHWO_OPTIMIZATION_DISABLE (1<<14)#define COMMON_SLICE_CHICKEN2 _MMIO(0x7014) +# define GEN9_PBE_COMPRESSED_HASH_SELECTION (1<<13)# define GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE (1<<12)# define GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION (1<<8)# define GEN8_CSC2_SBE_VUE_CACHE_CONSERVATIVE (1<<0)@@ -8071,6 +8072,7 @@ enum {#define GEN8_SAMPLER_POWER_BYPASS_DIS (1<<1)#define GEN9_HALF_SLICE_CHICKEN7 _MMIO(0xe194) +#define GEN9_SAMPLER_HASH_COMPRESSED_READ_ADDR (1<<8)#define GEN9_ENABLE_YV12_BUGFIX (1<<4)#define GEN9_ENABLE_GPGPU_PREEMPTION (1<<2)@@ -9371,4 +9373,8 @@ enum skl_power_gate {#define GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_SKL 0x67F1427F/* " " */ #define GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_BXT 0x5FF101FF/* " " */ +#define MMCD_MISC_CTRL _MMIO(0x4ddc)/* skl+ */ +#define MMCD_PCLA (1 << 31) +#define MMCD_HOTSPOT_EN (1 << 27) +#endif /* _I915_REG_H_ */ drivers/gpu/drm/i915/intel_engine_cs.c +13
-0@@ -812,6 +812,19 @@ static int gen9_init_workarounds(struct intel_engine_cs *engine)I915_WRITE (GAM_ECOCHK ,I915_READ (GAM_ECOCHK ) |ECOCHK_DIS_TLB ); +if (HAS_LLC (dev_priv)) { +/* WaCompressedResourceSamplerPbeMediaNewHashMode:skl,kbl +* +* Must match Display Engine. See +* WaCompressedResourceDisplayNewHashMode. +*/ +WA_SET_BIT_MASKED (COMMON_SLICE_CHICKEN2 , +GEN9_PBE_COMPRESSED_HASH_SELECTION ); +WA_SET_BIT_MASKED (GEN9_HALF_SLICE_CHICKEN7 , +GEN9_SAMPLER_HASH_COMPRESSED_READ_ADDR ); +WA_SET_BIT (MMCD_MISC_CTRL ,MMCD_PCLA |MMCD_HOTSPOT_EN ); + } +/* WaClearFlowControlGpgpuContextSave:skl,bxt,kbl,glk,cfl */ /* WaDisablePartialInstShootdown:skl,bxt,kbl,glk,cfl */ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,drivers/gpu/drm/i915/intel_pm.c +13
-14@@ -58,24 +58,23 @@static void gen9_init_clock_gating (struct drm_i915_private *dev_priv) { +if (HAS_LLC (dev_priv)) { +/* +* WaCompressedResourceDisplayNewHashMode:skl,kbl +* Display WA#0390: skl,kbl +* +* Must match Sampler, Pixel Back End, and Media. See +* WaCompressedResourceSamplerPbeMediaNewHashMode. +*/ +I915_WRITE (CHICKEN_PAR1_1 , +I915_READ (CHICKEN_PAR1_1 ) | +SKL_DE_COMPRESSED_HASH_MODE ); + } +/* See Bspec note for PSR2_CTL bit 31, Wa#828:skl,bxt,kbl,cfl */ I915_WRITE (CHICKEN_PAR1_1 ,I915_READ (CHICKEN_PAR1_1 ) |SKL_EDP_PSR_FIX_RDWRAP ); -/* -* Display WA#0390: skl,bxt,kbl,glk -* -* Must match Sampler, Pixel Back End, and Media -* (0xE194 bit 8, 0x7014 bit 13, 0x4DDC bits 27 and 31). -* -* Including bits outside the page in the hash would -* require 2 (or 4?) MiB alignment of resources. Just -* assume the defaul hashing mode which only uses bits -* within the page. -*/ -I915_WRITE (CHICKEN_PAR1_1 , -I915_READ (CHICKEN_PAR1_1 ) & ~SKL_RC_HASH_OUTSIDE ); -I915_WRITE (GEN8_CONFIG0 ,I915_READ (GEN8_CONFIG0 ) |GEN9_DEFAULT_FIXES );Signature · 1 line
2.13.0
Signature · 5 lines
Ben Widawsky, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 03/12] drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for CCS
Quoted 9 lines · Ben Widawsky
Quoted 5 lines · Ville Syrjälä
From: Ville Syrjälä <[email protected]>
Use the LLC/eLLC hotspot avoidance mode for CCS on LLC machines. This is reported to give better performance.
Not seeing in the diff how this only hits eLLC machines. Am I misreading when this is needed.
It's enabled on all LLC machines, not just those that have eLLC.
Quoted 128 lines
Quoted 125 lines
Testing has indicated that we don't need to enforce any massive 2 or 4 MiB alignment for all compressed resources even though there are still plenty of stale comments in the spec suggesting that we do.
We do need to make sure every hardware unit that deals with the compressed data uses the same hash mode.
Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>---
drivers/gpu/drm/i915/i915_reg.h | 8 +++++++- drivers/gpu/drm/i915/intel_engine_cs.c | 13 +++++++++++++ drivers/gpu/drm/i915/intel_pm.c | 27 +++++++++++++-------------- 3 files changed, 33 insertions(+), 15 deletions(-)drivers/gpu/drm/i915/i915_reg.h +7
-1@@ -6909,7 +6909,7 @@ enum {# define CHICKEN3_DGMG_DONE_FIX_DISABLE (1 << 2)#define CHICKEN_PAR1_1 _MMIO(0x42080) -#define SKL_RC_HASH_OUTSIDE (1 << 15) +#define SKL_DE_COMPRESSED_HASH_MODE (1 << 15)#define DPA_MASK_VBLANK_SRD (1 << 15)#define FORCE_ARB_IDLE_PLANES (1 << 14)#define SKL_EDP_PSR_FIX_RDWRAP (1 << 3)@@ -6982,6 +6982,7 @@ enum {# define GEN7_CSC1_RHWO_OPT_DISABLE_IN_RCC ((1<<10) | (1<<26))# define GEN9_RHWO_OPTIMIZATION_DISABLE (1<<14)#define COMMON_SLICE_CHICKEN2 _MMIO(0x7014) +# define GEN9_PBE_COMPRESSED_HASH_SELECTION (1<<13)# define GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE (1<<12)# define GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION (1<<8)# define GEN8_CSC2_SBE_VUE_CACHE_CONSERVATIVE (1<<0)@@ -8071,6 +8072,7 @@ enum {#define GEN8_SAMPLER_POWER_BYPASS_DIS (1<<1)#define GEN9_HALF_SLICE_CHICKEN7 _MMIO(0xe194) +#define GEN9_SAMPLER_HASH_COMPRESSED_READ_ADDR (1<<8)#define GEN9_ENABLE_YV12_BUGFIX (1<<4)#define GEN9_ENABLE_GPGPU_PREEMPTION (1<<2)@@ -9371,4 +9373,8 @@ enum skl_power_gate {#define GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_SKL 0x67F1427F/* " " */ #define GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_BXT 0x5FF101FF/* " " */ +#define MMCD_MISC_CTRL _MMIO(0x4ddc)/* skl+ */ +#define MMCD_PCLA (1 << 31) +#define MMCD_HOTSPOT_EN (1 << 27) +#endif /* _I915_REG_H_ */ drivers/gpu/drm/i915/intel_engine_cs.c +13
-0@@ -812,6 +812,19 @@ static int gen9_init_workarounds(struct intel_engine_cs *engine)I915_WRITE (GAM_ECOCHK ,I915_READ (GAM_ECOCHK ) |ECOCHK_DIS_TLB ); +if (HAS_LLC (dev_priv)) { +/* WaCompressedResourceSamplerPbeMediaNewHashMode:skl,kbl +* +* Must match Display Engine. See +* WaCompressedResourceDisplayNewHashMode. +*/ +WA_SET_BIT_MASKED (COMMON_SLICE_CHICKEN2 , +GEN9_PBE_COMPRESSED_HASH_SELECTION ); +WA_SET_BIT_MASKED (GEN9_HALF_SLICE_CHICKEN7 , +GEN9_SAMPLER_HASH_COMPRESSED_READ_ADDR ); +WA_SET_BIT (MMCD_MISC_CTRL ,MMCD_PCLA |MMCD_HOTSPOT_EN ); + } +/* WaClearFlowControlGpgpuContextSave:skl,bxt,kbl,glk,cfl */ /* WaDisablePartialInstShootdown:skl,bxt,kbl,glk,cfl */ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,drivers/gpu/drm/i915/intel_pm.c +13
-14@@ -58,24 +58,23 @@static void gen9_init_clock_gating (struct drm_i915_private *dev_priv) { +if (HAS_LLC (dev_priv)) { +/* +* WaCompressedResourceDisplayNewHashMode:skl,kbl +* Display WA#0390: skl,kbl +* +* Must match Sampler, Pixel Back End, and Media. See +* WaCompressedResourceSamplerPbeMediaNewHashMode. +*/ +I915_WRITE (CHICKEN_PAR1_1 , +I915_READ (CHICKEN_PAR1_1 ) | +SKL_DE_COMPRESSED_HASH_MODE ); + } +/* See Bspec note for PSR2_CTL bit 31, Wa#828:skl,bxt,kbl,cfl */ I915_WRITE (CHICKEN_PAR1_1 ,I915_READ (CHICKEN_PAR1_1 ) |SKL_EDP_PSR_FIX_RDWRAP ); -/* -* Display WA#0390: skl,bxt,kbl,glk -* -* Must match Sampler, Pixel Back End, and Media -* (0xE194 bit 8, 0x7014 bit 13, 0x4DDC bits 27 and 31). -* -* Including bits outside the page in the hash would -* require 2 (or 4?) MiB alignment of resources. Just -* assume the defaul hashing mode which only uses bits -* within the page. -*/ -I915_WRITE (CHICKEN_PAR1_1 , -I915_READ (CHICKEN_PAR1_1 ) & ~SKL_RC_HASH_OUTSIDE ); -I915_WRITE (GEN8_CONFIG0 ,I915_READ (GEN8_CONFIG0 ) |GEN9_DEFAULT_FIXES );Signature · 1 line
2.13.0Signature · 1 line
Ben Widawsky, Intel Open Source Technology Center
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 03/12] drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for CCS
In reply to [email protected]Quoted 16 lines · Ville Syrjälä
From: Ville Syrjälä <[email protected]>
Use the LLC/eLLC hotspot avoidance mode for CCS on LLC machines. This is reported to give better performance.
Testing has indicated that we don't need to enforce any massive 2 or 4 MiB alignment for all compressed resources even though there are still plenty of stale comments in the spec suggesting that we do.
We do need to make sure every hardware unit that deals with the compressed data uses the same hash mode.
Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>
[snip]
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 03/12] drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for CCS
Quoted 13 lines · Ben Widawsky
Quoted 12 lines · Ville Syrjälä
From: Ville Syrjälä <[email protected]>
Use the LLC/eLLC hotspot avoidance mode for CCS on LLC machines. This is reported to give better performance.
Testing has indicated that we don't need to enforce any massive 2 or 4 MiB alignment for all compressed resources even though there are still plenty of stale comments in the spec suggesting that we do.
We do need to make sure every hardware unit that deals with the compressed data uses the same hash mode.
Quoted 8 lines
Quoted 4 lines
Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>[snip]
Reviewed-by Ben Widawsky <[email protected]>
Thanks. Patch pushed to dinq.
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 4/12 drm/i915: Add a comment exlaining CCS hsub/vsub
In reply to [email protected]From: Ville Syrjälä <[email protected]>
Let's document why we claim hsub==8,vsub==16 for CCS even though the memory layout would suggest that we use 16x8 instead.
--- drivers/gpu/drm/i915/intel_display.c | 7 +++++++ 1 file changed, 7 insertions(+)
drivers/gpu/drm/i915/intel_display.c +7 -0
@@ -2484,6 +2484,13 @@ static unsigned int intel_fb_modifier_to_tiling(uint64_t fb_modifier) } } +/* +* 1 byte of CCS actually corresponds to 16x8 pixels on the main +* surface, and the memory layout for the CCS tile is 64x64 bytes. +* But since we're pretending the CCS tile is 128 bytes wide we +* adjust hsub/vsub here accordingly to 8x16 so that the +* bytes<->x/y conversions come out correct. +*/ static const struct drm_format_info ccs_formats[] = { { .format =DRM_FORMAT_XRGB8888 , .depth =24 , .num_planes =2 , .cpp = {4 ,1 , }, .hsub =8 , .vsub =16 , }, { .format =DRM_FORMAT_XBGR8888 , .depth =24 , .num_planes =2 , .cpp = {4 ,1 , }, .hsub =8 , .vsub =16 , },
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 5/12 drm/i915: Nuke a pointless unreachable()
In reply to [email protected]From: Ville Syrjälä <[email protected]>
The unreachable() is very much unreachable and the compiler knows that, so there's no point in having it.
--- drivers/gpu/drm/i915/intel_display.c | 2 -- 1 file changed, 2 deletions(-)
drivers/gpu/drm/i915/intel_display.c +0 -2
@@ -13078,8 +13078,6 @@ static bool intel_primary_plane_format_mod_supported(struct drm_plane *plane,return i965_mod_supported (format, modifier); else return i8xx_mod_supported(format, modifier); - -unreachable (); } static bool intel_cursor_plane_format_mod_supported(struct drm_plane *plane,
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 05/12] drm/i915: Nuke a pointless unreachable()
This was specifically requested by Emil.
Quoted 29 lines · Ville Syrjälä
From: Ville Syrjälä <[email protected]>
The
unreachable()is very much unreachable and the compiler knows that, so there's no point in having it.Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>---
drivers/gpu/drm/i915/intel_display.c| 2 --1 file changed, 2 deletions(-)drivers/gpu/drm/i915/intel_display.c +0
-2@@ -13078,8 +13078,6 @@ static bool intel_primary_plane_format_mod_supported(struct drm_plane *plane,return i965_mod_supported (format, modifier); else return i8xx_mod_supported(format, modifier); - -unreachable (); } static bool intel_cursor_plane_format_mod_supported(struct drm_plane *plane,Signature · 1 line
2.13.0
Signature · 5 lines
Ben Widawsky, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 05/12] drm/i915: Nuke a pointless unreachable()
Quoted 2 lines · Ben Widawsky
This was specifically requested by Emil.
Not quite, a gist is below.
Regardless, I do not want inspire lengthy discussions over a trivial suggestion. Please proceed as you gents prefer.
Thanks
Emil
a)Original code
if X
return A;
else if Y
return B;
else
return C;
return false;
b)My suggestion
if X
return A;
if Y
return B;
return C;
c)Ben's counter suggestion
if X
return A;
else if Y
return B;
else
return C;
unreachable();
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 6/12 drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
In reply to [email protected]From: Ville Syrjälä <[email protected]>
Y/Yf somehow dropped out from the SKL+ sprite modifier list. Add them in.
--- drivers/gpu/drm/i915/intel_sprite.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/gpu/drm/i915/intel_sprite.c +2 -0
@@ -1080,6 +1080,8 @@ static uint32_t skl_plane_formats[] = { };static const uint64_t skl_plane_format_modifiers[] = { + I915_FORMAT_MOD_Yf_TILED, +I915_FORMAT_MOD_Y_TILED ,I915_FORMAT_MOD_X_TILED ,DRM_FORMAT_MOD_LINEAR ,DRM_FORMAT_MOD_INVALID
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/12] drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Quoted 2 lines · On 24 August 2017 at 20:10, <[email protected]> wrote:
Y/Yf somehow dropped out from the SKL+ sprite modifier list. Add them in.
There's no 'somehow':
https://lists.freedesktop.org/archives/intel-gfx/2017-August/134932.html
I would prefer to not see this pushed whilst it doesn't actually work.
Cheers,
Daniel
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/12] drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Quoted 8 lines · Daniel Stone
Quoted 2 lines · On 24 August 2017 at 20:10, <[email protected]> wrote:
Y/Yf somehow dropped out from the SKL+ sprite modifier list. Add them in.
There's no 'somehow':
https://lists.freedesktop.org/archives/intel-gfx/2017-August/134932.htmlI would prefer to not see this pushed whilst it doesn't actually work.
Works fine here. Well, I should say it works just as well as it does for the primary plane. There are no plane specific checks in the wm/ddb code IIRC so if something is broken for sprites then it's most likely equally broken for the primary plane.
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/12] drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Hi,
Quoted 8 lines · Ville Syrjälä
Quoted 7 lines · Daniel Stone
Quoted 2 lines · On 24 August 2017 at 20:10, <[email protected]> wrote:
Y/Yf somehow dropped out from the SKL+ sprite modifier list. Add them in.
There's no 'somehow':
https://lists.freedesktop.org/archives/intel-gfx/2017-August/134932.html
Quoted 6 lines
Quoted 1 line
I would prefer to not see this pushed whilst it doesn't actually work.
Works fine here. Well, I should say it works just as well as it does for the primary plane. There are no plane specific checks in the wm/ddb code IIRC so if something is broken for sprites then it's most likely equally broken for the primary plane.
How did you test it?
The failure mode I observed was that the primary plane had a giant allocation, having previously had plain Y-tiling or Y-CCS enabled. After switching the primary to linear or X-tiled, trying to configure a 256x256 sprite plane with either Y-tiled or CCS failed, as it only had a DDB allocation of 31 blocks, where it needed 33. So it's not that there are plane-specific checks rejecting anything, it's that the allocations never got drawn up to give the sprite plane enough room to do anything other than X-tiled.
I saw this on both SKL (3200x1800 or 2560x1440) and APL (1920x1080).
Cheers,
Daniel
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 7/12 drm/i915: Clean up the sprite modifier checks
In reply to [email protected]From: Ville Syrjälä <[email protected]>
Split the g4x and snb cases into separate functions to match how we deal with all other platforms. Also sort the switch cases to match the format lists we've declared earlier, to ease comparisons.
--- drivers/gpu/drm/i915/intel_sprite.c | 48 ++++++++++++++++++++++--------------- 1 file changed, 29 insertions(+), 19 deletions(-)
drivers/gpu/drm/i915/intel_sprite.c +29 -19
@@ -1087,12 +1087,9 @@ static const uint64_t skl_plane_format_modifiers[] = { DRM_FORMAT_MOD_INVALID }; -static bool g4x_sprite_plane_format_mod_supported (struct drm_plane *plane, -uint32_t format, -uint64_t modifier) +static bool g4x_mod_supported (uint32_t format,uint64_t modifier) {switch (format) { -case DRM_FORMAT_XBGR8888 :case DRM_FORMAT_XRGB8888 :case DRM_FORMAT_YUYV :case DRM_FORMAT_YVYU :
@@ -1107,22 +1104,38 @@ static bool g4x_sprite_plane_format_mod_supported(struct drm_plane *plane, } } -static bool vlv_sprite_plane_format_mod_supported (struct drm_plane *plane, -uint32_t format, -uint64_t modifier) +static bool snb_mod_supported (uint32_t format,uint64_t modifier) {switch (format) { +case DRM_FORMAT_XRGB8888 : +case DRM_FORMAT_XBGR8888 :case DRM_FORMAT_YUYV :case DRM_FORMAT_YVYU :case DRM_FORMAT_UYVY :case DRM_FORMAT_VYUY : +if (modifier ==DRM_FORMAT_MOD_LINEAR || + modifier ==I915_FORMAT_MOD_X_TILED ) +return true; +/* fall through */ +default : +return false; + } +} + +static bool vlv_mod_supported (uint32_t format,uint64_t modifier) +{ +switch (format) {case DRM_FORMAT_RGB565 : -case DRM_FORMAT_XRGB8888 : +case DRM_FORMAT_ABGR8888 :case DRM_FORMAT_ARGB8888 : +case DRM_FORMAT_XBGR8888 : +case DRM_FORMAT_XRGB8888 :case DRM_FORMAT_XBGR2101010 :case DRM_FORMAT_ABGR2101010 : -case DRM_FORMAT_XBGR8888 : -case DRM_FORMAT_ABGR8888 : +case DRM_FORMAT_YUYV : +case DRM_FORMAT_YVYU : +case DRM_FORMAT_UYVY : +case DRM_FORMAT_VYUY :if (modifier ==DRM_FORMAT_MOD_LINEAR || modifier ==I915_FORMAT_MOD_X_TILED )return true;
@@ -1132,11 +1145,8 @@ static bool vlv_sprite_plane_format_mod_supported(struct drm_plane *plane, } } -static bool skl_sprite_plane_format_mod_supported (struct drm_plane *plane, -uint32_t format, -uint64_t modifier) +static bool skl_mod_supported (uint32_t format,uint64_t modifier) { -/* This is the same as primary plane since SKL has universal planes */ switch (format) {case DRM_FORMAT_XRGB8888 :case DRM_FORMAT_XBGR8888 :
@@ -1177,13 +1187,13 @@ static bool intel_sprite_plane_format_mod_supported(struct drm_plane *plane,return false;if (INTEL_GEN (dev_priv) >=9 ) -return skl_sprite_plane_format_mod_supported (plane, format, modifier); +return skl_mod_supported (format, modifier);else if (IS_VALLEYVIEW (dev_priv) ||IS_CHERRYVIEW (dev_priv)) -return vlv_sprite_plane_format_mod_supported (plane, format, modifier); +return vlv_mod_supported (format, modifier); +else if (INTEL_GEN (dev_priv) >=6 ) +return snb_mod_supported (format, modifier);else -return g4x_sprite_plane_format_mod_supported (plane, format, modifier); - -unreachable (); +return g4x_mod_supported (format, modifier); }static const struct drm_plane_funcs intel_sprite_plane_funcs = {
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 8/12 drm/i915: Add CCS capability for sprites
In reply to [email protected]From: Ville Syrjälä <[email protected]>
Allow sprites to scan out compressed framebuffers.
Since different platforms have a different set of planes that support CCS let's add a small helper to determine whether a specific plane supports CCS or not. Currently that information is spread around in many places, and not all the pieces of code even agree with each other.
In addition to allowing sprites to scan out compressed fbs, the other fix here is that we stop rejecting them on pipe C on CNL.
--- drivers/gpu/drm/i915/intel_display.c | 24 ++++-------------- drivers/gpu/drm/i915/intel_drv.h | 2 ++ drivers/gpu/drm/i915/intel_sprite.c | 49 +++++++++++++++++++++++++++--------- 3 files changed, 44 insertions(+), 31 deletions(-)
drivers/gpu/drm/i915/intel_display.c +5 -19
@@ -3120,6 +3120,7 @@ static int skl_check_nv12_aux_surface(struct intel_plane_state *plane_state)static int skl_check_ccs_aux_surface (struct intel_plane_state *plane_state) {struct intel_plane *plane =to_intel_plane (plane_state->base.plane); +struct drm_i915_private *dev_priv =to_i915 (plane->base.dev);struct intel_crtc *crtc =to_intel_crtc (plane_state->base.crtc);const struct drm_framebuffer *fb = plane_state->base.fb;int src_x = plane_state->base.src.x1 >>16 ;
@@ -3130,17 +3131,8 @@ static int skl_check_ccs_aux_surface(struct intel_plane_state *plane_state)int y = src_y / vsub;u32 offset; -switch (plane->id) { -case PLANE_PRIMARY : -case PLANE_SPRITE0 : -break ; -default : -DRM_DEBUG_KMS ("RC support only on plane 1 and 2\n" ); -return -EINVAL ; - } - -if (crtc->pipe ==PIPE_C ) { -DRM_DEBUG_KMS ("No RC support on pipe C\n" ); +if (!skl_plane_has_ccs (dev_priv, crtc->pipe, plane->id)) { +DRM_DEBUG_KMS ("No RC support on %s\n" , plane->base.name);return -EINVAL ; }
@@ -13279,17 +13271,11 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) primary->frontbuffer_bit =INTEL_FRONTBUFFER_PRIMARY (pipe); primary->check_plane = intel_check_primary_plane; -if (INTEL_GEN (dev_priv) >=10 ||IS_GEMINILAKE (dev_priv)) { +if (INTEL_GEN (dev_priv) >=9 ) { intel_primary_formats = skl_primary_formats; num_formats =ARRAY_SIZE (skl_primary_formats); - modifiers = skl_format_modifiers_ccs; - primary->update_plane = skylake_update_primary_plane; - primary->disable_plane = skylake_disable_primary_plane; - }else if (INTEL_GEN (dev_priv) >=9 ) { - intel_primary_formats = skl_primary_formats; - num_formats =ARRAY_SIZE (skl_primary_formats); -if (pipe <PIPE_C ) +if (skl_plane_has_ccs (dev_priv, pipe,PLANE_PRIMARY )) modifiers = skl_format_modifiers_ccs;else modifiers = skl_format_modifiers_noccs;
drivers/gpu/drm/i915/intel_drv.h +2 -0
@@ -1890,6 +1890,8 @@ int intel_sprite_set_colorkey(struct drm_device *dev, void *data, struct drm_file *file_priv);void intel_pipe_update_start (struct intel_crtc *crtc);void intel_pipe_update_end (struct intel_crtc *crtc); +bool skl_plane_has_ccs (struct drm_i915_private *dev_priv, +enum pipe pipe,enum plane_id plane_id);/* intel_tv.c */ void intel_tv_init (struct drm_i915_private *dev_priv);
drivers/gpu/drm/i915/intel_sprite.c +37 -12
80 lines not on this page · raw message
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 08/12] drm/i915: Add CCS capability for sprites
Quoted 10 lines · On Thu, 2017-08-24 at 22:10 +0300, [email protected] wrote:
From: Ville Syrjälä <[email protected]>
Allow sprites to scan out compressed framebuffers.
Since different platforms have a different set of planes that support CCS let's add a small helper to determine whether a specific plane supports CCS or not. Currently that information is spread around in many places, and not all the pieces of code even agree with each other.
Quoted 3 lines
In addition to allowing sprites to scan out compressed fbs, the other fix here is that we stop rejecting them on pipe C on CNL.
Unfortunately, this patch didn't apply cleanly on top of the latest drm-tip. Overall, the patch looks good to me.
Quoted 25 lines
Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>---
drivers/gpu/drm/i915/intel_display.c | 24 ++++-------------- drivers/gpu/drm/i915/intel_drv.h | 2 ++ drivers/gpu/drm/i915/intel_sprite.c | 49
+++++++++++++++++++++++++++---------3 files changed, 44 insertions(+), 31 deletions(-)drivers/gpu/drm/i915/intel_display.c +0
-0@@ -3120,6 +3120,7 @@ static int skl_check_nv12_aux_surface(struct intel_plane_state *plane_state)static int
skl_check_ccs_aux_surface(struct intel_plane_state *plane_state)
{
structintel_plane*plane = to_intel_plane(plane_state-Quoted 1 line
base.plane);
drivers/gpu/drm/i915/intel_display.c +1
-0+ struct drm_i915_private *dev_priv = to_i915(plane-
Quoted 1 line
Quoted 1 line
base.dev);
Seems that this addition is not really needed.
Quoted 163 lines
struct
intel_crtc*crtc = to_intel_crtc(plane_state-Quoted 1 line
base.crtc);
const struct
drm_framebuffer*fb =plane_state->base.fb; intsrc_x=plane_state->base.src.x1>> 16;drivers/gpu/drm/i915/intel_display.c +0
-0@@ -3130,17 +3131,8 @@ static int skl_check_ccs_aux_surface(struct intel_plane_state *plane_state) int y = src_y / vsub; u32 offset;drivers/gpu/drm/i915/intel_display.c +2
-11-switch (plane->id) { -case PLANE_PRIMARY : -case PLANE_SPRITE0 : -break ; -default : -DRM_DEBUG_KMS ("RC support only on plane 1 and 2\n" ); -return -EINVAL ; - } - -if (crtc->pipe ==PIPE_C ) { -DRM_DEBUG_KMS ("No RC support on pipe C\n" ); +if (!skl_plane_has_ccs (dev_priv, crtc->pipe, plane->id)) { + DRM_DEBUG_KMS("No RC support on %s\n", plane-Quoted 1 line
base.name);
return
-EINVAL;
}drivers/gpu/drm/i915/intel_display.c +0
-0@@ -13279,17 +13271,11 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) primary->frontbuffer_bit = INTEL_FRONTBUFFER_PRIMARY(pipe); primary->check_plane = intel_check_primary_plane;drivers/gpu/drm/i915/intel_display.c +1
-1-if (INTEL_GEN (dev_priv) >=10 ||IS_GEMINILAKE (dev_priv)) { +if (INTEL_GEN (dev_priv) >=9 ) {
intel_primary_formats=skl_primary_formats;num_formats=ARRAY_SIZE(skl_primary_formats);drivers/gpu/drm/i915/intel_display.c +0
-1- modifiers = skl_format_modifiers_ccs;drivers/gpu/drm/i915/intel_display.c +0
-1- primary->update_plane =
skylake_update_primary_plane;drivers/gpu/drm/i915/intel_display.c +0
-1- primary->disable_plane =
skylake_disable_primary_plane;drivers/gpu/drm/i915/intel_display.c +1
-4- }else if (INTEL_GEN (dev_priv) >=9 ) { - intel_primary_formats = skl_primary_formats; - num_formats =ARRAY_SIZE (skl_primary_formats); -if (pipe <PIPE_C ) + if (skl_plane_has_ccs(dev_priv, pipe,
PLANE_PRIMARY))
modifiers =skl_format_modifiers_ccs; else
modifiers =skl_format_modifiers_noccs;drivers/gpu/drm/i915/intel_drv.h +2
-0@@ -1890,6 +1890,8 @@ int intel_sprite_set_colorkey(struct drm_device *dev, void *data, struct drm_file *file_priv); void intel_pipe_update_start(struct intel_crtc *crtc); void intel_pipe_update_end(struct intel_crtc *crtc); +bool skl_plane_has_ccs (struct drm_i915_private *dev_priv, + enum pipe pipe, enum plane_id plane_id);
/*intel_tv.c*/
voidintel_tv_init(struct drm_i915_private *dev_priv);drivers/gpu/drm/i915/intel_sprite.c +0
-0@@ -1079,7 +1079,17 @@ static uint32_t skl_plane_formats[] = { DRM_FORMAT_VYUY, };drivers/gpu/drm/i915/intel_sprite.c +11
-1-static const uint64_t skl_plane_format_modifiers[] = { +static const uint64_t skl_plane_format_modifiers_noccs[] = { + I915_FORMAT_MOD_Yf_TILED, +I915_FORMAT_MOD_Y_TILED , +I915_FORMAT_MOD_X_TILED , +DRM_FORMAT_MOD_LINEAR , +DRM_FORMAT_MOD_INVALID +}; + +static const uint64_t skl_plane_format_modifiers_ccs[] = { + I915_FORMAT_MOD_Yf_TILED_CCS, +I915_FORMAT_MOD_Y_TILED_CCS ,
I915_FORMAT_MOD_Yf_TILED,
I915_FORMAT_MOD_Y_TILED,
I915_FORMAT_MOD_X_TILED,drivers/gpu/drm/i915/intel_sprite.c +0
-0@@ -1152,6 +1162,10 @@ static bool skl_mod_supported(uint32_t format, uint64_t modifier)case
DRM_FORMAT_XBGR8888:
caseDRM_FORMAT_ARGB8888:
caseDRM_FORMAT_ABGR8888:drivers/gpu/drm/i915/intel_sprite.c +4
-0+if (modifier == I915_FORMAT_MOD_Yf_TILED_CCS || + modifier == I915_FORMAT_MOD_Y_TILED_CCS) +return true; +/* fall through */ case
DRM_FORMAT_RGB565:
caseDRM_FORMAT_XRGB2101010:
caseDRM_FORMAT_XBGR2101010:drivers/gpu/drm/i915/intel_sprite.c +0
-0@@ -1207,6 +1221,23 @@ static const struct drm_plane_funcs intel_sprite_plane_funcs = {.format_mod_supported =
intel_sprite_plane_format_mod_supported,
};drivers/gpu/drm/i915/intel_sprite.c +17
-0+bool skl_plane_has_ccs (struct drm_i915_private *dev_priv, + enum pipe pipe, enum plane_id plane_id) +{ +if (plane_id ==PLANE_CURSOR ) +return false; + +if (INTEL_GEN (dev_priv) >=10 ) +return true; + +if (IS_GEMINILAKE (dev_priv)) +return pipe !=PIPE_C ; + +return pipe !=PIPE_C && + (plane_id ==PLANE_PRIMARY || + plane_id == PLANE_SPRITE0); +} +struct
intel_plane*intel_sprite_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe, int plane)drivers/gpu/drm/i915/intel_sprite.c +0
-0@@ -1233,7 +1264,7 @@ intel_sprite_plane_create(structdrm_i915_private *dev_priv, } intel_plane->base.state = &state->base;drivers/gpu/drm/i915/intel_sprite.c +1
-1-if (INTEL_GEN (dev_priv) >=10 ) { +if (INTEL_GEN (dev_priv) >=9 ) {
intel_plane->can_scale= true;
state->scaler_id= -1;drivers/gpu/drm/i915/intel_sprite.c +0
-0@@ -1242,17 +1273,11 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv,
plane_formats=skl_plane_formats;num_plane_formats=ARRAY_SIZE(skl_plane_formats);drivers/gpu/drm/i915/intel_sprite.c +0
-7- modifiers = skl_plane_format_modifiers; - }else if (INTEL_GEN (dev_priv) >=9 ) { - intel_plane->can_scale = true; - state->scaler_id =-1 ; - - intel_plane->update_plane = skl_update_plane; - intel_plane->disable_plane = skl_disable_plane;drivers/gpu/drm/i915/intel_sprite.c +5
-3- plane_formats = skl_plane_formats; - num_plane_formats =ARRAY_SIZE (skl_plane_formats); - modifiers = skl_plane_format_modifiers; +if (skl_plane_has_ccs (dev_priv, pipe,PLANE_SPRITE0 + plane)) + modifiers = skl_plane_format_modifiers_ccs; +else + modifiers =
skl_plane_format_modifiers_noccs;
} else if (IS_VALLEYVIEW(dev_priv)||IS_CHERRYVIEW(dev_priv)) {
intel_plane->can_scale= false;
intel_plane->max_downscale= 1;
Signature · 6 lines
Mika Kahola - Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 08/12] drm/i915: Add CCS capability for sprites
Hi Mika,
Quoted 9 lines · Mika Kahola
Quoted 8 lines · On Thu, 2017-08-24 at 22:10 +0300, [email protected] wrote:
Allow sprites to scan out compressed framebuffers.
Since different platforms have a different set of planes that support CCS let's add a small helper to determine whether a specific plane supports CCS or not. Currently that information is spread around in many places, and not all the pieces of code even agree with each other.
Quoted 5 lines
Quoted 3 lines
In addition to allowing sprites to scan out compressed fbs, the other fix here is that we stop rejecting them on pipe C on CNL.
Unfortunately, this patch didn't apply cleanly on top of the latest drm-tip. Overall, the patch looks good to me.
Did you manage to test this? When I tried, the DDB/watermark allocation was too conservative for sprites, and never allowed enough blocks to be able to use anything but linear or X-tiled on sprite planes. The allocation was almost entirely taken up by the primary plane, even if that itself was later set to linear or X-tiled.
I observed this on SKL with both 3200x1800 and 2560x1440 output sizes (and primary plane sizes; no scaling), as well as APL with 1920x1080.
Cheers,
Daniel
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 08/12] drm/i915: Add CCS capability for sprites
Quoted 19 lines · Daniel Stone
Hi Mika,
Quoted 15 lines · Mika Kahola
Quoted 12 lines · On Thu, 2017-08-24 at 22:10 +0300, [email protected] wrote:
Allow sprites to scan out compressed framebuffers.
Since different platforms have a different set of planes that support CCS let's add a small helper to determine whether a specific plane supports CCS or not. Currently that information is spread around in many places, and not all the pieces of code even agree with each other.
In addition to allowing sprites to scan out compressed fbs, the other fix here is that we stop rejecting them on pipe C on CNL.
Quoted 7 lines
Quoted 2 lines
Unfortunately, this patch didn't apply cleanly on top of the latest drm-tip. Overall, the patch looks good to me.
Did you manage to test this? When I tried, the DDB/watermark allocation was too conservative for sprites, and never allowed enough blocks to be able to use anything but linear or X-tiled on sprite planes. The allocation was almost entirely taken up by the primary plane, even if that itself was later set to linear or X-tiled.
I tried to test this but unfortunately this didn't apply on top of drm- tip. The whole series dates back to August so a lot have changed since then.
I would need this patch for my CNL work so I would like to see this merged.
Quoted 6 lines
I observed this on SKL with both 3200x1800 and 2560x1440 output sizes (and primary plane sizes; no scaling), as well as APL with 1920x1080.
Cheers,
Daniel
Signature · 6 lines
Mika Kahola - Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 08/12] drm/i915: Add CCS capability for sprites
Hi,
Quoted 7 lines · Mika Kahola
Quoted 5 lines · Daniel Stone
Did you manage to test this? When I tried, the DDB/watermark allocation was too conservative for sprites, and never allowed enough blocks to be able to use anything but linear or X-tiled on sprite planes. The allocation was almost entirely taken up by the primary plane, even if that itself was later set to linear or X-tiled.
Quoted 6 lines
I tried to test this but unfortunately this didn't apply on top of drm- tip. The whole series dates back to August so a lot have changed since then.
I would need this patch for my CNL work so I would like to see this merged.
I would like to see this merged, but I would also like to see it ever working first. :) When I tried it at the time, even a 256x256 plane could not use Y-tiling or CCS, due to DDB exhaustion.
Cheers,
Daniel
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 08/12] drm/i915: Add CCS capability for sprites
Quoted 14 lines · Daniel Stone
Hi,
Quoted 11 lines · Mika Kahola
Quoted 5 lines · Daniel Stone
Did you manage to test this? When I tried, the DDB/watermark allocation was too conservative for sprites, and never allowed enough blocks to be able to use anything but linear or X-tiled on sprite planes. The allocation was almost entirely taken up by the primary plane, even if that itself was later set to linear or X-tiled.
I tried to test this but unfortunately this didn't apply on top of drm- tip. The whole series dates back to August so a lot have changed since then.
Quoted 6 lines
Quoted 2 lines
I would need this patch for my CNL work so I would like to see this merged.
I would like to see this merged, but I would also like to see it ever working first. :) When I tried it at the time, even a 256x256 plane could not use Y-tiling or CCS, due to DDB exhaustion.
It worked just fine when I tried it a few months ago.
22:45 < vsyrjala> my skl seems perfectly happy here with two 2560x1080 planes at same time scanning out yccs
22:47 < danvet> I think daniels had a huge screen or something like that 22:47 < vsyrjala> let me try my 4k
22:54 < vsyrjala> ok, so now i hit a wm limit
23:01 < vsyrjala> so two 3k x 2x planes both scanning out yccs works 23:02 < vsyrjala> 3.5k x 2k fails
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 9/12 drm/i915: Allow up to 32KB stride on SKL+ "sprites"
In reply to [email protected]From: Ville Syrjälä <[email protected]>
SKL+ "sprites" no longer have 16KB max stride limit that earlier platforms had. Bump up the limit to 32KB.
--- drivers/gpu/drm/i915/intel_sprite.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_sprite.c +2 -1
@@ -786,6 +786,7 @@ intel_check_sprite_plane(struct intel_plane *plane,struct drm_rect *src = &state->base.src;struct drm_rect *dst = &state->base.dst;const struct drm_rect *clip = &state->clip; +int max_stride =INTEL_GEN (dev_priv) >=9 ?32768 :16384 ;int hscale, vscale;int max_scale, min_scale;bool can_scale;
@@ -806,7 +807,7 @@ intel_check_sprite_plane(struct intel_plane *plane, }/* FIXME check all gen limits */ -if (fb->width <3 || fb->height <3 || fb->pitches[0 ] >16384 ) { +if (fb->width <3 || fb->height <3 || fb->pitches[0 ] > max_stride) {DRM_DEBUG_KMS ("Unsuitable framebuffer for plane\n" );return -EINVAL ; }
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 10/12 drm: Fix modifiers_property kernel doc
In reply to [email protected]From: Ville Syrjälä <[email protected]>
The member is called 'modifiers_property' instead of 'modifiers'. Adjust the kernel docs to match.
--- include/drm/drm_mode_config.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
include/drm/drm_mode_config.h +1 -1
@@ -758,7 +758,7 @@ struct drm_mode_config {bool allow_fb_modifiers;/** -* @modifiers: Plane property to list support modifier/format +* @modifiers_property: Plane property to list support modifier/format * combination. */ struct drm_property *modifiers_property;
Signature · 6 lines
2.13.0 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 10/12] drm: Fix modifiers_property kernel doc
Quoted 10 lines · On Thu, Aug 24, 2017 at 10:10:58PM +0300, [email protected] wrote:
From: Ville Syrjälä <[email protected]>
The member is called '
modifiers_property' instead of 'modifiers'. Adjust the kernel docs to match.Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>
Pls push to drm-misc-next (imo no need to for this to land in 4.14). -Daniel
Quoted 24 lines
--- include/drm/drm_mode_config.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)include/drm/drm_mode_config.h +1
-1@@ -758,7 +758,7 @@ struct drm_mode_config {bool allow_fb_modifiers;/** -* @modifiers: Plane property to list support modifier/format +* @modifiers_property: Plane property to list support modifier/format * combination. */ struct drm_property *modifiers_property;Signature · 6 lines
2.13.0 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH 11/12 drm: Check that the plane supports the request format+modifier combo
In reply to [email protected]From: Ville Syrjälä <[email protected]>
Currently we only check that the plane supports the pixel format of the fb we're about to feed to it. Extend it to check also the modifier, and more specifically that the combination of the format and modifier is supported.
--- drivers/gpu/drm/drm_atomic.c | 8 +++++--- drivers/gpu/drm/drm_crtc.c | 8 +++++--- drivers/gpu/drm/drm_crtc_internal.h | 4 ++-- drivers/gpu/drm/drm_plane.c | 31 +++++++++++++++++++++++++------ 4 files changed, 37 insertions(+), 14 deletions(-)
drivers/gpu/drm/drm_atomic.c +5 -3
@@ -884,12 +884,14 @@ static int drm_atomic_plane_check(struct drm_plane *plane, }/* Check whether this plane supports the fb pixel format. */ - ret =drm_plane_check_pixel_format (plane, state->fb->format->format); + ret =drm_plane_check_pixel_format (plane, state->fb->format->format, + state->fb->modifier);if (ret) {struct drm_format_name_buf format_name; -DRM_DEBUG_ATOMIC ("Invalid pixel format %s\n" , +DRM_DEBUG_ATOMIC ("Invalid pixel format %s, modifier 0x%llx\n" ,drm_get_format_name (state->fb->format->format, - &format_name)); + &format_name), + state->fb->modifier);return ret; }
drivers/gpu/drm/drm_crtc.c +5 -3
@@ -625,12 +625,14 @@ int drm_mode_setcrtc(struct drm_device *dev, void *data,*/ if (!crtc->primary->format_default) { ret =drm_plane_check_pixel_format (crtc->primary, - fb->format->format); + fb->format->format, + fb->modifier);if (ret) {struct drm_format_name_buf format_name; -DRM_DEBUG_KMS ("Invalid pixel format %s\n" , +DRM_DEBUG_KMS ("Invalid pixel format %s, modifier 0x%llx\n" ,drm_get_format_name (fb->format->format, - &format_name)); + &format_name), + fb->modifier); goto out; } }
drivers/gpu/drm/drm_crtc_internal.h +2 -2
@@ -194,8 +194,8 @@ int drm_mode_atomic_ioctl(struct drm_device *dev,/* drm_plane.c */ int drm_plane_register_all (struct drm_device *dev);void drm_plane_unregister_all (struct drm_device *dev); -int drm_plane_check_pixel_format (const struct drm_plane *plane, -u32 format); +int drm_plane_check_pixel_format (struct drm_plane *plane, +u32 format,u64 modifier);/* drm_bridge.c */ void drm_bridge_detach (struct drm_bridge *bridge);
drivers/gpu/drm/drm_plane.c +25 -6
@@ -555,16 +555,33 @@ int drm_mode_getplane(struct drm_device *dev, void *data,return 0 ; } -int drm_plane_check_pixel_format (const struct drm_plane *plane,u32 format) +int drm_plane_check_pixel_format (struct drm_plane *plane, +u32 format,u64 modifier) {unsigned int i;for (i =0 ; i < plane->format_count; i++) {if (format == plane->format_types[i]) -return 0 ; +break ; + } +if (i == plane->format_count) +return -EINVAL ; + +if (!plane->modifier_count) +return 0 ; + +for (i =0 ; i < plane->modifier_count; i++) { +if (modifier == plane->modifiers[i]) +break ; } +if (i == plane->modifier_count) +return -EINVAL ; -return -EINVAL ; +if (plane->funcs->format_mod_supported && + !plane->funcs->format_mod_supported (plane, format, modifier)) +return -EINVAL ; + +return 0 ; }/*
@@ -609,12 +626,14 @@ static int __setplane_internal(struct drm_plane *plane, }/* Check whether this plane supports the fb pixel format. */ - ret =drm_plane_check_pixel_format (plane, fb->format->format); + ret =drm_plane_check_pixel_format (plane, fb->format->format, + fb->modifier);if (ret) {struct drm_format_name_buf format_name; -DRM_DEBUG_KMS ("Invalid pixel format %s\n" , +DRM_DEBUG_KMS ("Invalid pixel format %s, modifier 0x%llx\n" ,drm_get_format_name (fb->format->format, - &format_name)); + &format_name), + fb->modifier); goto out; }
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 11/12] drm: Check that the plane supports the request format+modifier combo
Quoted 7 lines · On Thu, Aug 24, 2017 at 10:10:59PM +0300, [email protected] wrote:
From: Ville Syrjälä <[email protected]>
Currently we only check that the plane supports the pixel format of the fb we're about to feed to it. Extend it to check also the modifier, and more specifically that the combination of the format and modifier is supported.
Quoted 5 lines
Cc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>Signed-off-by Ville Syrjälä <[email protected]>
I think Daniel Stone is on the hook to augment kms_ccs to properly test this all in igt. Would be nice to add the corresponding Testcase: lines when that's done.
With the test coverage gap addressed:
Quoted 132 lines
--- drivers/gpu/drm/drm_atomic.c | 8 +++++--- drivers/gpu/drm/drm_crtc.c | 8 +++++--- drivers/gpu/drm/drm_crtc_internal.h | 4 ++-- drivers/gpu/drm/drm_plane.c | 31 +++++++++++++++++++++++++------ 4 files changed, 37 insertions(+), 14 deletions(-)drivers/gpu/drm/drm_atomic.c +5
-3@@ -884,12 +884,14 @@ static int drm_atomic_plane_check(struct drm_plane *plane, }/* Check whether this plane supports the fb pixel format. */ - ret =drm_plane_check_pixel_format (plane, state->fb->format->format); + ret =drm_plane_check_pixel_format (plane, state->fb->format->format, + state->fb->modifier);if (ret) {struct drm_format_name_buf format_name; -DRM_DEBUG_ATOMIC ("Invalid pixel format %s\n" , +DRM_DEBUG_ATOMIC ("Invalid pixel format %s, modifier 0x%llx\n" ,drm_get_format_name (state->fb->format->format, - &format_name)); + &format_name), + state->fb->modifier);return ret; }drivers/gpu/drm/drm_crtc.c +5
-3@@ -625,12 +625,14 @@ int drm_mode_setcrtc(struct drm_device *dev, void *data,*/ if (!crtc->primary->format_default) { ret =drm_plane_check_pixel_format (crtc->primary, - fb->format->format); + fb->format->format, + fb->modifier);if (ret) {struct drm_format_name_buf format_name; -DRM_DEBUG_KMS ("Invalid pixel format %s\n" , +DRM_DEBUG_KMS ("Invalid pixel format %s, modifier 0x%llx\n" ,drm_get_format_name (fb->format->format, - &format_name)); + &format_name), + fb->modifier); goto out; } }drivers/gpu/drm/drm_crtc_internal.h +2
-2@@ -194,8 +194,8 @@ int drm_mode_atomic_ioctl(struct drm_device *dev,/* drm_plane.c */ int drm_plane_register_all (struct drm_device *dev);void drm_plane_unregister_all (struct drm_device *dev); -int drm_plane_check_pixel_format (const struct drm_plane *plane, -u32 format); +int drm_plane_check_pixel_format (struct drm_plane *plane, +u32 format,u64 modifier);/* drm_bridge.c */ void drm_bridge_detach (struct drm_bridge *bridge);drivers/gpu/drm/drm_plane.c +25
-6@@ -555,16 +555,33 @@ int drm_mode_getplane(struct drm_device *dev, void *data,return 0 ; } -int drm_plane_check_pixel_format (const struct drm_plane *plane,u32 format) +int drm_plane_check_pixel_format (struct drm_plane *plane, +u32 format,u64 modifier) {unsigned int i;for (i =0 ; i < plane->format_count; i++) {if (format == plane->format_types[i]) -return 0 ; +break ; + } +if (i == plane->format_count) +return -EINVAL ; + +if (!plane->modifier_count) +return 0 ; + +for (i =0 ; i < plane->modifier_count; i++) { +if (modifier == plane->modifiers[i]) +break ; } +if (i == plane->modifier_count) +return -EINVAL ; -return -EINVAL ; +if (plane->funcs->format_mod_supported && + !plane->funcs->format_mod_supported (plane, format, modifier)) +return -EINVAL ; + +return 0 ; }/* @@ -609,12 +626,14 @@ static int __setplane_internal(struct drm_plane *plane, }/* Check whether this plane supports the fb pixel format. */ - ret =drm_plane_check_pixel_format (plane, fb->format->format); + ret =drm_plane_check_pixel_format (plane, fb->format->format, + fb->modifier);if (ret) {struct drm_format_name_buf format_name; -DRM_DEBUG_KMS ("Invalid pixel format %s\n" , +DRM_DEBUG_KMS ("Invalid pixel format %s, modifier 0x%llx\n" ,drm_get_format_name (fb->format->format, - &format_name)); + &format_name), + fb->modifier); goto out; }Signature · 6 lines
2.13.0 _______________________________________________ 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 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH 12/12 drm/i915: Remove the pipe/plane ID checks from skl_check_ccs_aux_surface()
In reply to [email protected]From: Ville Syrjälä <[email protected]>
The core now checks that the plane supports the fb's format+modifier combination, so we can drop the related checks from skl_check_ccs_aux_surface(). These checks were specific to SKL/KBL/BXT anyway.
--- drivers/gpu/drm/i915/intel_display.c | 8 -------- 1 file changed, 8 deletions(-)
drivers/gpu/drm/i915/intel_display.c +0 -8
@@ -3119,9 +3119,6 @@ static int skl_check_nv12_aux_surface(struct intel_plane_state *plane_state)static int skl_check_ccs_aux_surface (struct intel_plane_state *plane_state) { -struct intel_plane *plane =to_intel_plane (plane_state->base.plane); -struct drm_i915_private *dev_priv =to_i915 (plane->base.dev); -struct intel_crtc *crtc =to_intel_crtc (plane_state->base.crtc);const struct drm_framebuffer *fb = plane_state->base.fb;int src_x = plane_state->base.src.x1 >>16 ;int src_y = plane_state->base.src.y1 >>16 ;
@@ -3131,11 +3128,6 @@ static int skl_check_ccs_aux_surface(struct intel_plane_state *plane_state)int y = src_y / vsub;u32 offset; -if (!skl_plane_has_ccs (dev_priv, crtc->pipe, plane->id)) { -DRM_DEBUG_KMS ("No RC support on %s\n" , plane->base.name); -return -EINVAL ; - } -if (plane_state->base.rotation & ~(DRM_MODE_ROTATE_0 |DRM_MODE_ROTATE_180 )) {DRM_DEBUG_KMS ("RC support only with 0/180 degree rotation %x\n" , plane_state->base.rotation);
Signature · 6 lines
2.13.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Patchwork
CI✓ Fi.CI.BAT: success for drm/i915: Fix up the CCS code
In reply to [email protected]== Series Details ==
== Series Details ==
Series: drm/i915: Fix up the CCS code
URL : https://patchwork.freedesktop.org/series/29308/
State : success
== Summary ==
Series 29308v1 drm/i915: Fix up the CCS code
https://patchwork.freedesktop.org/api/1.0/series/29308/revisions/1/mbox/
Test kms_flip:
Subgroup basic-flip-vs-dpms:
dmesg-warn -> PASS (fi-kbl-7560u) fdo#102392 +2fdo#102392 https://bugs.freedesktop.org/show_bug.cgi?id=102392
fi-bdw-5557u total:279 pass:268 dwarn:0 dfail:0 fail:0 skip:11 time:453s fi-bdw-gvtdvm total:279 pass:265 dwarn:0 dfail:0 fail:0 skip:14 time:435s fi-blb-e6850 total:279 pass:224 dwarn:1 dfail:0 fail:0 skip:54 time:357s fi-bsw-n3050 total:279 pass:243 dwarn:0 dfail:0 fail:0 skip:36 time:550s fi-bwr-2160 total:279 pass:184 dwarn:0 dfail:0 fail:0 skip:95 time:254s fi-bxt-j4205 total:279 pass:260 dwarn:0 dfail:0 fail:0 skip:19 time:528s fi-byt-j1900 total:279 pass:254 dwarn:1 dfail:0 fail:0 skip:24 time:522s fi-byt-n2820 total:279 pass:251 dwarn:0 dfail:0 fail:0 skip:28 time:517s fi-elk-e7500 total:279 pass:230 dwarn:0 dfail:0 fail:0 skip:49 time:436s fi-glk-2a total:279 pass:260 dwarn:0 dfail:0 fail:0 skip:19 time:609s fi-hsw-4770 total:279 pass:263 dwarn:0 dfail:0 fail:0 skip:16 time:447s fi-hsw-4770r total:279 pass:263 dwarn:0 dfail:0 fail:0 skip:16 time:425s fi-ilk-650 total:279 pass:229 dwarn:0 dfail:0 fail:0 skip:50 time:422s fi-ivb-3520m total:279 pass:261 dwarn:0 dfail:0 fail:0 skip:18 time:504s fi-ivb-3770 total:279 pass:261 dwarn:0 dfail:0 fail:0 skip:18 time:473s fi-kbl-7500u total:279 pass:261 dwarn:0 dfail:0 fail:0 skip:18 time:480s fi-kbl-7560u total:210 pass:205 dwarn:1 dfail:0 fail:0 skip:3 fi-kbl-r total:279 pass:261 dwarn:0 dfail:0 fail:0 skip:18 time:603s fi-pnv-d510 total:279 pass:223 dwarn:1 dfail:0 fail:0 skip:55 time:535s fi-skl-6260u total:279 pass:269 dwarn:0 dfail:0 fail:0 skip:10 time:468s fi-skl-6700k total:279 pass:261 dwarn:0 dfail:0 fail:0 skip:18 time:477s fi-skl-6770hq total:279 pass:269 dwarn:0 dfail:0 fail:0 skip:10 time:485s fi-skl-gvtdvm total:279 pass:266 dwarn:0 dfail:0 fail:0 skip:13 time:441s fi-snb-2520m total:279 pass:251 dwarn:0 dfail:0 fail:0 skip:28 time:542s fi-snb-2600 total:279 pass:250 dwarn:0 dfail:0 fail:0 skip:29 time:405s
650ffdaee44b186cd59cbaf893d4177bc0a78aa1 drm-tip: 2017y-08m-24d-14h-29m-45s UTC integration manifest 98e730858092 drm/i915: Remove the pipe/plane ID checks from skl_check_ccs_aux_surface() 0a4a48834e23 drm: Check that the plane supports the request format+modifier combo 027ccca2d8e8 drm: Fix modifiers_property kernel doc9444cc1bd712 drm/i915: Allow up to 32KB stride on SKL+ "sprites"878da5624fc8 drm/i915: Add CCS capability for sprites5b4e7bd649b2 drm/i915: Clean up the sprite modifier checksae345cacbe26 drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites3ad385207cb8 drm/i915: Nuke a pointless unreachable()91e2802bfaed drm/i915: Add a comment exlaining CCS hsub/vsub25f698d358cf drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for CCS 52144a3d7fed drm/i915: Skip fence alignemnt check for the CCS plane2628f6f3d9c8 drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear offset
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_5487/
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Patchwork
CI✓ Fi.CI.IGT: success for drm/i915: Fix up the CCS code
In reply to [email protected]== Series Details ==
== Series Details ==
Series: drm/i915: Fix up the CCS code
URL : https://patchwork.freedesktop.org/series/29308/
State : success
== Summary ==
Test kms_setmode:
Subgroup basic:
fail -> PASS (shard-hsw) fdo#99912fdo#99912 https://bugs.freedesktop.org/show_bug.cgi?id=99912
shard-hsw total:2230 pass:1232 dwarn:0 dfail:0 fail:16 skip:982 time:9457s
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_5487/shards.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 11 lines · On Thu, Aug 24, 2017 at 10:10:48PM +0300, [email protected] wrote:
From: Ville Syrjälä <[email protected]>
Looks like we ended up with a stale version of my CCS code in dinq. This series contains the remainder in smaller chunks. I also ended up adding a bunch of extra cleanup etc. on top.
The most important thing we need to get in is the change to the fb->offsets[] interpretation since that's ABI territory. The hash mode apparently doesn't require the nasty virtual address alignment tricks so that shouldn't have any ABI issues after all.
Quoted 6 lines
Entire series available here:
git://github.com/vsyrjala/linux.gitccs_fixesCc Ben Widawsky <[email protected]>Cc Jason Ekstrand <[email protected]>Cc Daniel Stone <[email protected]>
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
Thanks, Daniel
Quoted 38 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offset
drm/i915: Skip fence alignemnt check for the CCS plane drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCS
drm/i915: Add a comment exlaining CCS hsub/vsub
drm/i915: Nuke a pointlessunreachable()
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites drm/i915: Clean up the sprite modifier checks
drm/i915: Add CCS capability for sprites
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
drm: Fixmodifiers_propertykernel doc
drm: Check that the plane supports the request format+modifier combo drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()drivers/gpu/drm/drm_atomic.c | 8 +- drivers/gpu/drm/drm_crtc.c | 8 +- drivers/gpu/drm/drm_crtc_internal.h | 4 +- drivers/gpu/drm/drm_plane.c | 31 +++++-- drivers/gpu/drm/i915/i915_reg.h | 8 +- drivers/gpu/drm/i915/intel_display.c | 145 +++++++++++++++++---------------- drivers/gpu/drm/i915/intel_drv.h | 2 + drivers/gpu/drm/i915/intel_engine_cs.c | 13 +++ drivers/gpu/drm/i915/intel_pm.c | 27 +++--- drivers/gpu/drm/i915/intel_sprite.c | 102 +++++++++++++++-------- include/drm/drm_mode_config.h | 2 +- 11 files changed, 217 insertions(+), 133 deletions(-)Signature · 6 lines
2.13.0 _______________________________________________ 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
Daniel Stone
NAKHi Daniel,
Quoted 4 lines · Daniel Vetter
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
My opinion below; tl;dr is that I don't think most of them are super-critical. Ville obviously has a far stronger opinion than me on the shape of the code, so I'm fine with this series, which seems to mostly be a merge back of the delta between whatever Ville's latest branch was, and whatever the last patchset Ben sent out was.
Quoted 3 lines
Quoted 3 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offset
This should land into -fixes. I trust Ville that it has no UABI impact, but seems like something to be very consistent on.
Quoted 1 line
Quoted 1 line
drm/i915: Skip fence alignemnt check for the CCS plane
Not sure if this is -fixes material really, just a cleanup?
Quoted 2 lines
Quoted 2 lines
drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCS
Not -fixes, performance optimisation.
Quoted 1 line
Quoted 1 line
drm/i915: Add a comment exlaining CCS hsub/vsub
Seems harmless to land to -fixes.
Quoted 1 line
Quoted 1 line
drm/i915: Nuke a pointless
unreachable()
Ditto.
Quoted 1 line
Quoted 1 line
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Per my previous reply, NAK to landing at all, since DDB/WM allocation seems too broken for it to work.
Quoted 1 line
Quoted 1 line
drm/i915: Clean up the sprite modifier checks
Fine with this, but doesn't seem like -fixes material.
Quoted 1 line
Quoted 1 line
drm/i915: Add CCS capability for sprites
NAK, same reason as Y/Yf.
Quoted 1 line
Quoted 1 line
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
Again doesn't seem like -fixes necessarily?
Quoted 1 line
Quoted 1 line
drm: Fix
modifiers_propertykernel doc
Good for -fixes.
Quoted 1 line
Quoted 1 line
drm: Check that the plane supports the request format+modifier combo
Good for core (not Intel) -fixes.
Quoted 2 lines
Quoted 2 lines
drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()
Seems fine but probably not -fixes material; land in Intel after a merge?
Cheers,
Daniel
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 14 lines · Daniel Stone
Hi Daniel,
Quoted 4 lines · Daniel Vetter
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
My opinion below; tl;dr is that I don't think most of them are super-critical. Ville obviously has a far stronger opinion than me on the shape of the code, so I'm fine with this series, which seems to mostly be a merge back of the delta between whatever Ville's latest branch was, and whatever the last patchset Ben sent out was.
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offsetThis should land into -fixes. I trust Ville that it has no UABI impact, but seems like something to be very consistent on.
It does change the uabi. That's the whole point. What was merged doesn't agree with what userspace wants. So this we want in definitely so that we don't end up exposing the wrong uabi in any released kernel.
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Skip fence alignemnt check for the CCS plane
Not sure if this is -fixes material really, just a cleanup?
It makes the kernel less likely to reject the fb entirely. So without this userspace has to be rather careful where it places the aux surface. I would include this as well.
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCSNot -fixes, performance optimisation.
We hope. It does change the layout of the compressed data though so if our testcases try to generate compressed data with the CPU it'll not go well if the test assumes the wrong hash mode. I would include this as well so that we don't end up in any kind of a mess later when we try to change it.
So the patches were more or less sorted in priority order, and we want at least 01,02 and maybe 03.
Quoted 41 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Add a comment exlaining CCS hsub/vsub
Seems harmless to land to -fixes.
Quoted 1 line
Quoted 1 line
drm/i915: Nuke a pointless
unreachable()Ditto.
Quoted 1 line
Quoted 1 line
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Per my previous reply, NAK to landing at all, since DDB/WM allocation seems too broken for it to work.
Quoted 1 line
Quoted 1 line
drm/i915: Clean up the sprite modifier checks
Fine with this, but doesn't seem like -fixes material.
Quoted 1 line
Quoted 1 line
drm/i915: Add CCS capability for sprites
NAK, same reason as Y/Yf.
Quoted 1 line
Quoted 1 line
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
Again doesn't seem like -fixes necessarily?
Quoted 1 line
Quoted 1 line
drm: Fix
modifiers_propertykernel docGood for -fixes.
Quoted 1 line
Quoted 1 line
drm: Check that the plane supports the request format+modifier combo
Good for core (not Intel) -fixes.
Quoted 2 lines
Quoted 2 lines
drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()Seems fine but probably not -fixes material; land in Intel after a merge?
Cheers,
Daniel
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Jani Nikula
maintainerQuoted 41 lines · Ville Syrjälä
Quoted 14 lines · Daniel Stone
Hi Daniel,
Quoted 4 lines · Daniel Vetter
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
My opinion below; tl;dr is that I don't think most of them are super-critical. Ville obviously has a far stronger opinion than me on the shape of the code, so I'm fine with this series, which seems to mostly be a merge back of the delta between whatever Ville's latest branch was, and whatever the last patchset Ben sent out was.
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offsetThis should land into -fixes. I trust Ville that it has no UABI impact, but seems like something to be very consistent on.
It does change the uabi. That's the whole point. What was merged doesn't agree with what userspace wants. So this we want in definitely so that we don't end up exposing the wrong uabi in any released kernel.
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Skip fence alignemnt check for the CCS plane
Not sure if this is -fixes material really, just a cleanup?
It makes the kernel less likely to reject the fb entirely. So without this userspace has to be rather careful where it places the aux surface. I would include this as well.
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCSNot -fixes, performance optimisation.
Quoted 8 lines
We hope. It does change the layout of the compressed data though so if our testcases try to generate compressed data with the CPU it'll not go well if the test assumes the wrong hash mode. I would include this as well so that we don't end up in any kind of a mess later when we try to change it.
So the patches were more or less sorted in priority order, and we want at least 01,02 and maybe 03.
When you decide what to apply, please *please* add the appropriate
Fixes: tags for the ones you want to show up in v4.14.
BR,
Jani.
Quoted 42 lines
Quoted 42 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Add a comment exlaining CCS hsub/vsub
Seems harmless to land to -fixes.
Quoted 1 line
Quoted 1 line
drm/i915: Nuke a pointless
unreachable()Ditto.
Quoted 1 line
Quoted 1 line
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Per my previous reply, NAK to landing at all, since DDB/WM allocation seems too broken for it to work.
Quoted 1 line
Quoted 1 line
drm/i915: Clean up the sprite modifier checks
Fine with this, but doesn't seem like -fixes material.
Quoted 1 line
Quoted 1 line
drm/i915: Add CCS capability for sprites
NAK, same reason as Y/Yf.
Quoted 1 line
Quoted 1 line
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
Again doesn't seem like -fixes necessarily?
Quoted 1 line
Quoted 1 line
drm: Fix
modifiers_propertykernel docGood for -fixes.
Quoted 1 line
Quoted 1 line
drm: Check that the plane supports the request format+modifier combo
Good for core (not Intel) -fixes.
Quoted 2 lines
Quoted 2 lines
drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()Seems fine but probably not -fixes material; land in Intel after a merge?
Cheers,
Daniel
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 48 lines · Jani Nikula
Quoted 47 lines · Ville Syrjälä
Quoted 14 lines · Daniel Stone
Hi Daniel,
Quoted 4 lines · Daniel Vetter
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
My opinion below; tl;dr is that I don't think most of them are super-critical. Ville obviously has a far stronger opinion than me on the shape of the code, so I'm fine with this series, which seems to mostly be a merge back of the delta between whatever Ville's latest branch was, and whatever the last patchset Ben sent out was.
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offsetThis should land into -fixes. I trust Ville that it has no UABI impact, but seems like something to be very consistent on.
It does change the uabi. That's the whole point. What was merged doesn't agree with what userspace wants. So this we want in definitely so that we don't end up exposing the wrong uabi in any released kernel.
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Skip fence alignemnt check for the CCS plane
Not sure if this is -fixes material really, just a cleanup?
It makes the kernel less likely to reject the fb entirely. So without this userspace has to be rather careful where it places the aux surface. I would include this as well.
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCSNot -fixes, performance optimisation.
We hope. It does change the layout of the compressed data though so if our testcases try to generate compressed data with the CPU it'll not go well if the test assumes the wrong hash mode. I would include this as well so that we don't end up in any kind of a mess later when we try to change it.
Quoted 5 lines
Quoted 2 lines
So the patches were more or less sorted in priority order, and we want at least 01,02 and maybe 03.
When you decide what to apply, please *please* add the appropriate
Fixes: tags for the ones you want to show up in v4.14.
I just pushed 01 and 02 to dinq with the approriage Fixes: tags. I'd still prefer to get 03 in as well, but that would need an r-b/ack.
Quoted 50 lines
BR,
Jani.Quoted 42 lines
Quoted 42 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Add a comment exlaining CCS hsub/vsub
Seems harmless to land to -fixes.
Quoted 1 line
Quoted 1 line
drm/i915: Nuke a pointless
unreachable()Ditto.
Quoted 1 line
Quoted 1 line
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Per my previous reply, NAK to landing at all, since DDB/WM allocation seems too broken for it to work.
Quoted 1 line
Quoted 1 line
drm/i915: Clean up the sprite modifier checks
Fine with this, but doesn't seem like -fixes material.
Quoted 1 line
Quoted 1 line
drm/i915: Add CCS capability for sprites
NAK, same reason as Y/Yf.
Quoted 1 line
Quoted 1 line
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
Again doesn't seem like -fixes necessarily?
Quoted 1 line
Quoted 1 line
drm: Fix
modifiers_propertykernel docGood for -fixes.
Quoted 1 line
Quoted 1 line
drm: Check that the plane supports the request format+modifier combo
Good for core (not Intel) -fixes.
Quoted 2 lines
Quoted 2 lines
drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()Seems fine but probably not -fixes material; land in Intel after a merge?
Cheers,
DanielSignature · 1 line
Jani Nikula, Intel Open Source Technology Center
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Hi Ville,
Quoted 52 lines · Ville Syrjälä
Quoted 51 lines · Jani Nikula
Quoted 49 lines · Ville Syrjälä
Quoted 14 lines · Daniel Stone
Hi Daniel,
Quoted 4 lines · Daniel Vetter
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
My opinion below; tl;dr is that I don't think most of them are super-critical. Ville obviously has a far stronger opinion than me on the shape of the code, so I'm fine with this series, which seems to mostly be a merge back of the delta between whatever Ville's latest branch was, and whatever the last patchset Ben sent out was.
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offsetThis should land into -fixes. I trust Ville that it has no UABI impact, but seems like something to be very consistent on.
It does change the uabi. That's the whole point. What was merged doesn't agree with what userspace wants. So this we want in definitely so that we don't end up exposing the wrong uabi in any released kernel.
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Skip fence alignemnt check for the CCS plane
Not sure if this is -fixes material really, just a cleanup?
It makes the kernel less likely to reject the fb entirely. So without this userspace has to be rather careful where it places the aux surface. I would include this as well.
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCSNot -fixes, performance optimisation.
We hope. It does change the layout of the compressed data though so if our testcases try to generate compressed data with the CPU it'll not go well if the test assumes the wrong hash mode. I would include this as well so that we don't end up in any kind of a mess later when we try to change it.
So the patches were more or less sorted in priority order, and we want at least 01,02 and maybe 03.
Quoted 54 lines
Quoted 2 lines
When you decide what to apply, please *please* add the appropriate
Fixes: tags for the ones you want to show up in v4.14.I just pushed 01 and 02 to dinq with the approriage Fixes: tags. I'd still prefer to get 03 in as well, but that would need an r-b/ack.
Quoted 47 lines
BR,
Jani.Quoted 42 lines
Quoted 42 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Add a comment exlaining CCS hsub/vsub
Seems harmless to land to -fixes.
Quoted 1 line
Quoted 1 line
drm/i915: Nuke a pointless
unreachable()Ditto.
Quoted 1 line
Quoted 1 line
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Per my previous reply, NAK to landing at all, since DDB/WM allocation seems too broken for it to work.
Quoted 1 line
Quoted 1 line
drm/i915: Clean up the sprite modifier checks
Fine with this, but doesn't seem like -fixes material.
Quoted 1 line
Quoted 1 line
drm/i915: Add CCS capability for sprites
NAK, same reason as Y/Yf.
Quoted 1 line
Quoted 1 line
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
Again doesn't seem like -fixes necessarily?
Quoted 1 line
Quoted 1 line
drm: Fix
modifiers_propertykernel docGood for -fixes.
Quoted 1 line
Quoted 1 line
drm: Check that the plane supports the request format+modifier combo
Good for core (not Intel) -fixes.
Quoted 2 lines
Quoted 2 lines
drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()Seems fine but probably not -fixes material; land in Intel after a merge?
Cheers,
Daniel
Should I wait any more of this for drm-intel-next-fixes?
Otherwise I will move with the pull request.
Thanks,
Rodrigo.
Quoted 11 lines
Quoted 3 lines
--
Jani Nikula, Intel Open Source Technology Center--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 6 lines
Rodrigo Vivi Blog: http://blog.vivi.eng.br _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 106 lines · Rodrigo Vivi
Hi Ville,
Quoted 102 lines · Ville Syrjälä
Quoted 48 lines · Jani Nikula
Quoted 47 lines · Ville Syrjälä
Quoted 14 lines · Daniel Stone
Hi Daniel,
Quoted 4 lines · Daniel Vetter
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
My opinion below; tl;dr is that I don't think most of them are super-critical. Ville obviously has a far stronger opinion than me on the shape of the code, so I'm fine with this series, which seems to mostly be a merge back of the delta between whatever Ville's latest branch was, and whatever the last patchset Ben sent out was.
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offsetThis should land into -fixes. I trust Ville that it has no UABI impact, but seems like something to be very consistent on.
It does change the uabi. That's the whole point. What was merged doesn't agree with what userspace wants. So this we want in definitely so that we don't end up exposing the wrong uabi in any released kernel.
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Skip fence alignemnt check for the CCS plane
Not sure if this is -fixes material really, just a cleanup?
It makes the kernel less likely to reject the fb entirely. So without this userspace has to be rather careful where it places the aux surface. I would include this as well.
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCSNot -fixes, performance optimisation.
We hope. It does change the layout of the compressed data though so if our testcases try to generate compressed data with the CPU it'll not go well if the test assumes the wrong hash mode. I would include this as well so that we don't end up in any kind of a mess later when we try to change it.
Quoted 5 lines
Quoted 2 lines
So the patches were more or less sorted in priority order, and we want at least 01,02 and maybe 03.
When you decide what to apply, please *please* add the appropriate
Fixes: tags for the ones you want to show up in v4.14.I just pushed 01 and 02 to dinq with the approriage Fixes: tags. I'd still prefer to get 03 in as well, but that would need an r-b/ack.
Quoted 43 lines
BR,
Jani.Quoted 38 lines
Quoted 38 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Add a comment exlaining CCS hsub/vsub
Seems harmless to land to -fixes.
Quoted 1 line
Quoted 1 line
drm/i915: Nuke a pointless
unreachable()Ditto.
Quoted 1 line
Quoted 1 line
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Per my previous reply, NAK to landing at all, since DDB/WM allocation seems too broken for it to work.
Quoted 1 line
Quoted 1 line
drm/i915: Clean up the sprite modifier checks
Fine with this, but doesn't seem like -fixes material.
Quoted 1 line
Quoted 1 line
drm/i915: Add CCS capability for sprites
NAK, same reason as Y/Yf.
Quoted 1 line
Quoted 1 line
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
Again doesn't seem like -fixes necessarily?
Quoted 1 line
Quoted 1 line
drm: Fix
modifiers_propertykernel docGood for -fixes.
Quoted 1 line
Quoted 1 line
drm: Check that the plane supports the request format+modifier combo
Good for core (not Intel) -fixes.
Quoted 2 lines
Quoted 2 lines
drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Seems fine but probably not -fixes material; land in Intel after a merge?
Cheers,
DanielShould I wait any more of this for drm-intel-next-fixes?
Otherwise I will move with the pull request.
Go ahead with the pull request. We should be able to live with just the first two patches for now.
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 84 lines · Ville Syrjälä
Quoted 83 lines · Rodrigo Vivi
Hi Ville,
Quoted 79 lines · Ville Syrjälä
Quoted 48 lines · Jani Nikula
Quoted 47 lines · Ville Syrjälä
Quoted 14 lines · Daniel Stone
Hi Daniel,
Quoted 4 lines · Daniel Vetter
Which of these do we need to cherry-pick over to -next-fixes? There's no annotations about that. If the answer is "most" I'm leaning towards disabling CCS for 4.14, minimal set would be ideal (and first in the patch series).
My opinion below; tl;dr is that I don't think most of them are super-critical. Ville obviously has a far stronger opinion than me on the shape of the code, so I'm fine with this series, which seems to mostly be a merge back of the delta between whatever Ville's latest branch was, and whatever the last patchset Ben sent out was.
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
Ville Syrjälä (12):
drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear
offsetThis should land into -fixes. I trust Ville that it has no UABI impact, but seems like something to be very consistent on.
It does change the uabi. That's the whole point. What was merged doesn't agree with what userspace wants. So this we want in definitely so that we don't end up exposing the wrong uabi in any released kernel.
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Skip fence alignemnt check for the CCS plane
Not sure if this is -fixes material really, just a cleanup?
It makes the kernel less likely to reject the fb entirely. So without this userspace has to be rather careful where it places the aux surface. I would include this as well.
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Switch over to the LLC/eLLC hotspot avoidance hash mode for
CCSNot -fixes, performance optimisation.
We hope. It does change the layout of the compressed data though so if our testcases try to generate compressed data with the CPU it'll not go well if the test assumes the wrong hash mode. I would include this as well so that we don't end up in any kind of a mess later when we try to change it.
Quoted 5 lines
Quoted 2 lines
So the patches were more or less sorted in priority order, and we want at least 01,02 and maybe 03.
When you decide what to apply, please *please* add the appropriate
Fixes: tags for the ones you want to show up in v4.14.I just pushed 01 and 02 to dinq with the approriage Fixes: tags. I'd still prefer to get 03 in as well, but that would need an r-b/ack.
Quoted 20 lines
BR,
Jani.Quoted 15 lines
Quoted 15 lines
Quoted 3 lines
Quoted 3 lines
drm/i915: Add a comment exlaining CCS hsub/vsub
Seems harmless to land to -fixes.
Quoted 1 line
Quoted 1 line
drm/i915: Nuke a pointless
unreachable()Ditto.
Quoted 1 line
Quoted 1 line
drm/i915: Add the missing Y/Yf modifiers for SKL+ sprites
Per my previous reply, NAK to landing at all, since DDB/WM allocation seems too broken for it to work.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 1 line
Quoted 1 line
drm/i915: Clean up the sprite modifier checks
Fine with this, but doesn't seem like -fixes material.
Quoted 1 line
Quoted 1 line
drm/i915: Add CCS capability for sprites
NAK, same reason as Y/Yf.
What's the status for CCS_E on the overlay plane?
thanks,
Kristian
Quoted 28 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 2 lines
Quoted 2 lines
drm/i915: Allow up to 32KB stride on SKL+ "sprites"
Again doesn't seem like -fixes necessarily?
Quoted 1 line
Quoted 1 line
drm: Fix
modifiers_propertykernel docGood for -fixes.
Quoted 1 line
Quoted 1 line
drm: Check that the plane supports the request format+modifier combo
Good for core (not Intel) -fixes.
Quoted 2 lines
Quoted 2 lines
drm/i915: Remove the pipe/plane ID checks from
skl_check_ccs_aux_surface()Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Seems fine but probably not -fixes material; land in Intel after a merge?
Cheers,
DanielShould I wait any more of this for drm-intel-next-fixes?
Otherwise I will move with the pull request.
Go ahead with the pull request. We should be able to live with just the first two patches for now.
Quoted 7 lines
--
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx