RK3288 Gamma LUT
Subsystems (5)
v3 cover RK3288 Gamma LUT
Let's support Gamma LUT configuration on RK3288 SoCs.
In order to do so, this series adds a new and optional address resource.
A separate address resource is required because on this RK3288, the LUT address is after the MMU address, which is requested by the iommu driver. This prevents the DRM driver from requesting an entire register space.
The current implementation works for RGB 10-bit tables, as that is what seems to work on RK3288.
This has been tested on a Rock2 Square board, using 'modetest' tool (modetest now supports GAMMA_LUT property), with legacy and atomic APIs.
In addition, I've tested it with Jacopo's modified kmsxx [1], See the rcar-du color management series for more information [2].
[1] https://jmondi.org/cgit/kmsxx/
[2] https://lkml.org/lkml/2019/9/6/490
Thanks,
Eze
Changes from v2:
* revert Sean Paul's patch, in order to use
atomic_commit_tail hook.
* add RFC/patch for color management on resume.
Changes from v1:
* drop explicit linear LUT after finding a proper way to disable gamma correction.
* avoid setting gamma is the CRTC is not active.
* s/int/unsigned int as suggested by Jacopo.
* only enable color management and set gamma size if gamma LUT is supported, suggested by Doug.
* drop the reg-names usage, and instead just use indexed reg specifiers, suggested by Doug.
Changes from RFC:
* Request (an optional) address resource for the LUT.
* Add devicetree changes.
* Drop support for RK3399, which doesn't seem to work out of the box and needs more research.
* Support pass-thru setting when GAMMA_LUT is NULL.
* Add a check for the gamma size, as suggested by Ilia.
* Move gamma setting to atomic_commit_tail, as pointed out by Jacopo/Laurent, is the correct way.
Ezequiel Garcia (5):
Revert "drm/rockchip: Use drm_atomic_helper_commit_tail_rpm"
dt-bindings: display: rockchip: document VOP gamma LUT address drm/rockchip: Add optional support for CRTC gamma LUT
ARM: dts: rockchip: Add RK3288 VOP gamma LUT address
RFC: drm/atomic-helper: Reapply color transformation after resume
.../display/rockchip/rockchip-vop.txt | 6 +- arch/arm/boot/dts/rk3288.dtsi | 4 +- drivers/gpu/drm/drm_atomic_helper.c | 12 ++ drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 24 +++- drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 114 ++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 7 ++ drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 2 + 7 files changed, 165 insertions(+), 4 deletions(-)
Signature · 1 line
2.22.0
PATCH v3 1/5 Revert "drm/rockchip: Use drm_atomic_helper_commit_tail_rpm"
This reverts commit c87fb38df19da3362a0e20df1aad852100995ead, which conflicts with adding driver-specific behavior in atomic_commit_tail().
No functional changes expected on this commit, but just preparation for the upcoming gamma LUT support.
---
Changes from v2:
* New patch.
--- drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +20 -1
@@ -105,8 +105,27 @@ rockchip_user_fb_create(struct drm_device *dev, struct drm_file *file_priv,return ERR_PTR (ret); } +static void +rockchip_atomic_helper_commit_tail_rpm (struct drm_atomic_state *old_state) +{ +struct drm_device *dev = old_state->dev; + +drm_atomic_helper_commit_modeset_disables (dev, old_state); + +drm_atomic_helper_commit_modeset_enables (dev, old_state); + +drm_atomic_helper_commit_planes (dev, old_state, +DRM_PLANE_COMMIT_ACTIVE_ONLY ); + +drm_atomic_helper_commit_hw_done (old_state); + +drm_atomic_helper_wait_for_vblanks (dev, old_state); + +drm_atomic_helper_cleanup_planes (dev, old_state); +} +static const struct drm_mode_config_helper_funcs rockchip_mode_config_helpers = { - .atomic_commit_tail = drm_atomic_helper_commit_tail_rpm, + .atomic_commit_tail = rockchip_atomic_helper_commit_tail_rpm, };static const struct drm_mode_config_funcs rockchip_drm_mode_config_funcs = {
Signature · 1 line
2.22.0
PATCH v3 2/5 dt-bindings: display: rockchip: document VOP gamma LUT address
In reply to Ezequiel GarciaAdd the register specifier description for an optional gamma LUT address.
---
Changes from v2:
* None.
Changes from v1:
* Drop reg-names, suggested by Doug.
--- .../devicetree/bindings/display/rockchip/rockchip-vop.txt | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/display/rockchip/rockchip-vop.txt +5 -1
@@ -20,6 +20,10 @@ Required properties: "rockchip,rk3228-vop"; "rockchip,rk3328-vop"; +- reg: Must contain one entry corresponding to the base address and length + of the register space. Can optionally contain a second entry + corresponding to the CRTC gamma LUT address. + - interrupts: should contain a list of all VOP IP block interrupts in the order: VSYNC, LCD_SYSTEM. The interrupt specifier format depends on the interrupt controller used.
@@ -48,7 +52,7 @@ Example: SoC specific DT entry: vopb: vopb@ff930000 { compatible = "rockchip,rk3288-vop"; - reg = <0xff930000 0x19c>; + reg = <0x0 0xff930000 0x0 0x19c>, <0x0 0xff931000 0x0 0x1000>; interrupts = <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>; clocks = <&cru ACLK_VOP0>, <&cru DCLK_VOP0>, <&cru HCLK_VOP0>; clock-names = "aclk_vop", "dclk_vop", "hclk_vop";
Signature · 1 line
2.22.0
PATCH v3 3/5 drm/rockchip: Add optional support for CRTC gamma LUT
In reply to Ezequiel GarciaAdd an optional CRTC gamma LUT support, and enable it on RK3288. This is currently enabled via a separate address resource, which needs to be specified in the devicetree.
The address resource is required because on some SoCs, such as RK3288, the LUT address is after the MMU address, and the latter is supported by a different driver. This prevents the DRM driver from requesting an entire register space.
The current implementation works for RGB 10-bit tables, as that is what seems to work on RK3288.
---
Changes from v2:
* None.
Changes from v1:
* drop explicit linear LUT after finding a proper way to disable gamma correction.
* avoid setting gamma is the CRTC is not active.
* s/int/unsigned int as suggested by Jacopo.
* only enable color management and set gamma size if gamma LUT is supported, suggested by Doug.
* drop the reg-names usage, and instead just use indexed reg specifiers, suggested by Doug.
Changes from RFC:
* Request (an optional) address resource for the LUT.
* Drop support for RK3399, which doesn't seem to work out of the box and needs more research.
* Support pass-thru setting when GAMMA_LUT is NULL.
* Add a check for the gamma size, as suggested by Ilia.
* Move gamma setting to atomic_commit_tail, as pointed out by Jacopo/Laurent, is the correct way.
--- drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 3 + drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 114 ++++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 7 ++ drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 2 + 4 files changed, 126 insertions(+)
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +3 -0
@@ -17,6 +17,7 @@#include "rockchip_drm_drv.h" #include "rockchip_drm_fb.h" #include "rockchip_drm_gem.h" +#include "rockchip_drm_vop.h" static const struct drm_framebuffer_funcs rockchip_drm_fb_funcs = { .destroy = drm_gem_fb_destroy,
@@ -112,6 +113,8 @@ rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state)drm_atomic_helper_commit_modeset_disables (dev, old_state); +rockchip_drm_vop_gamma_set (old_state); +drm_atomic_helper_commit_modeset_enables (dev, old_state); drm_atomic_helper_commit_planes(dev, old_state,
drivers/gpu/drm/rockchip/rockchip_drm_vop.c +114 -0
150 lines not on this page · raw message
drivers/gpu/drm/rockchip/rockchip_drm_vop.h +7 -0
@@ -7,6 +7,8 @@#ifndef _ROCKCHIP_DRM_VOP_H#define _ROCKCHIP_DRM_VOP_H +#include <drm/drm_atomic.h> +/* * major: IP major version, used for IP structure * minor: big feature change under same structure
@@ -67,6 +69,7 @@ struct vop_common {struct vop_reg dither_down_mode;struct vop_reg dither_down_en;struct vop_reg dither_up; +struct vop_reg dsp_lut_en;struct vop_reg gate_en;struct vop_reg mmu_en;struct vop_reg out_mode;
@@ -170,6 +173,7 @@ struct vop_data {const struct vop_win_yuv2yuv_data *win_yuv2yuv;const struct vop_win_data *win;unsigned int win_size; +unsigned int lut_size;#define VOP_FEATURE_OUTPUT_RGB10 BIT(0)#define VOP_FEATURE_INTERNAL_RGB BIT(1)
@@ -373,4 +377,7 @@ static inline int scl_vop_cal_lb_mode(int width, bool is_yuv) }extern const struct component_ops vop_component_ops; + +void rockchip_drm_vop_gamma_set (struct drm_atomic_state *state); +#endif /* _ROCKCHIP_DRM_VOP_H */
drivers/gpu/drm/rockchip/rockchip_vop_reg.c +2 -0
@@ -598,6 +598,7 @@ static const struct vop_common rk3288_common = { .dither_down_en = VOP_REG(RK3288_DSP_CTRL1, 0x1, 2), .pre_dither_down = VOP_REG(RK3288_DSP_CTRL1, 0x1, 1), .dither_up = VOP_REG(RK3288_DSP_CTRL1, 0x1, 6), + .dsp_lut_en = VOP_REG(RK3288_DSP_CTRL1, 0x1, 0), .data_blank = VOP_REG(RK3288_DSP_CTRL0, 0x1, 19), .dsp_blank = VOP_REG(RK3288_DSP_CTRL0, 0x3, 18), .out_mode = VOP_REG(RK3288_DSP_CTRL0, 0xf, 0),
@@ -646,6 +647,7 @@ static const struct vop_data rk3288_vop = { .output = &rk3288_output, .win = rk3288_vop_win_data, .win_size = ARRAY_SIZE(rk3288_vop_win_data), + .lut_size = 1024, };static const int rk3368_vop_intrs[] = {
Signature · 1 line
2.22.0
Re: [PATCH v3 3/5] drm/rockchip: Add optional support for CRTC gamma LUT
Quoted 57 lines · Ezequiel Garcia
Add an optional CRTC gamma LUT support, and enable it on RK3288. This is currently enabled via a separate address resource, which needs to be specified in the devicetree.
The address resource is required because on some SoCs, such as RK3288, the LUT address is after the MMU address, and the latter is supported by a different driver. This prevents the DRM driver from requesting an entire register space.
The current implementation works for RGB 10-bit tables, as that is what seems to work on RK3288.
Signed-off-by Ezequiel Garcia <[email protected]>Reviewed-by Douglas Anderson <[email protected]>Reviewed-by Jacopo Mondi <[email protected]>---
Changes from v2:
* None.Changes from v1:
* drop explicit linear LUT after finding a proper way to disable gamma correction.
* avoid setting gamma is the CRTC is not active.
* s/int/unsigned int as suggested by Jacopo.
* only enable color management and set gamma size if gamma LUT is supported, suggested by Doug.
* drop the reg-names usage, and instead just use indexed reg specifiers, suggested by Doug.Changes from RFC:
* Request (an optional) address resource for the LUT.
* Drop support for RK3399, which doesn't seem to work out of the box and needs more research.
* Support pass-thru setting whenGAMMA_LUTis NULL.
* Add a check for the gamma size, as suggested by Ilia.
* Move gamma setting toatomic_commit_tail, as pointed out by Jacopo/Laurent, is the correct way.--- drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 3 + drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 114 ++++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 7 ++ drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 2 + 4 files changed, 126 insertions(+)drivers/gpu/drm/rockchip/rockchip_drm_fb.c +1
-0@@ -17,6 +17,7 @@#include "rockchip_drm_drv.h" #include "rockchip_drm_fb.h" #include "rockchip_drm_gem.h" +#include "rockchip_drm_vop.h" static const struct drm_framebuffer_funcs rockchip_drm_fb_funcs = { .destroy = drm_gem_fb_destroy,
drivers/gpu/drm/rockchip/rockchip_drm_fb.c rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state) · line 117
Quoted 5 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +2
-0@@ -112,6 +113,8 @@ rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state)drm_atomic_helper_commit_modeset_disables (dev, old_state); +rockchip_drm_vop_gamma_set (old_state); +
Instead of duplicating the commit_tail helper, could you just implement .atomic_begin() and call this from there? I think the only hitch is if you need this to be completed before crtc->atomic_enable(), at which point you might need to call it from vop_crtc_atomic_enable() and then detect that in atomic_begin()
That would prevent the revert in patch 1 and keep rockchip on the helper train.
Sean
Quoted 223 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +0
-0drm_atomic_helper_commit_modeset_enables (dev, old_state); drm_atomic_helper_commit_planes(dev, old_state,drivers/gpu/drm/rockchip/rockchip_drm_vop.c +114
-0@@ -141,6 +141,7 @@ struct vop {uint32_t *regsbak;void __iomem *regs; +void __iomem *lut_regs;/* physical map length of vop register */ uint32_t len;@@ -1193,6 +1194,102 @@ static void vop_wait_for_irq_handler(struct vop *vop)synchronize_irq (vop->irq); } +static bool vop_dsp_lut_is_enable (struct vop *vop) +{ +return vop_read_reg (vop,0 , &vop->data->common->dsp_lut_en); +} + +static void vop_crtc_write_gamma_lut (struct vop *vop,struct drm_crtc *crtc) +{ +struct drm_color_lut *lut = crtc->state->gamma_lut->data; +unsigned int i; + +for (i =0 ; i < crtc->gamma_size; i++) { +u32 word; + + word = (drm_color_lut_extract (lut[i].red,10 ) <<20 ) | + (drm_color_lut_extract (lut[i].green,10 ) <<10 ) | +drm_color_lut_extract (lut[i].blue,10 ); +writel (word, vop->lut_regs + i *4 ); + } +} + +static void vop_crtc_gamma_set (struct vop *vop,struct drm_crtc *crtc, +struct drm_crtc_state *old_state) +{ +unsigned int idle; +int ret; + +/* +* In order to write the LUT to the internal RAM memory, +* we need to first make sure the dsp_lut_en bit is cleared. +*/ +spin_lock (&vop->reg_lock); +VOP_REG_SET (vop, common, dsp_lut_en,0 ); +vop_cfg_done (vop); +spin_unlock (&vop->reg_lock); + +/* +* If the CRTC is not active, dsp_lut_en will not get cleared. +* Apparently we still need to do the above step to for +* gamma correction to be disabled. +*/ +if (!crtc->state->active) +return ; + + ret =readx_poll_timeout (vop_dsp_lut_is_enable, vop, + idle, !idle,5 ,30 *1000 ); +if (ret) { +DRM_DEV_ERROR (vop->dev,"display LUT RAM enable timeout!\n" ); +return ; + } + +spin_lock (&vop->reg_lock); + +if (crtc->state->gamma_lut && + (!old_state->gamma_lut || (crtc->state->gamma_lut->base.id != + old_state->gamma_lut->base.id))) +vop_crtc_write_gamma_lut (vop, crtc); + +VOP_REG_SET (vop, common, dsp_lut_en,1 ); +vop_cfg_done (vop); +spin_unlock (&vop->reg_lock); +} + +static int vop_crtc_atomic_check (struct drm_crtc *crtc, +struct drm_crtc_state *crtc_state) +{ +struct vop *vop =to_vop (crtc); + +if (vop->lut_regs && crtc_state->color_mgmt_changed && + crtc_state->gamma_lut) { +unsigned int len; + + len =drm_color_lut_size (crtc_state->gamma_lut); +if (len != crtc->gamma_size) { +DRM_DEBUG_KMS ("Invalid LUT size; got %d, expected %d\n" , + len, crtc->gamma_size); +return -EINVAL ; + } + } + +return 0 ; +} + +void rockchip_drm_vop_gamma_set (struct drm_atomic_state *state) +{ +struct drm_crtc_state *old_crtc_state; +struct drm_crtc *crtc; +unsigned int i; + +for_each_old_crtc_in_state (state, crtc, old_crtc_state, i) { +struct vop *vop =to_vop (crtc); + +if (vop->lut_regs && crtc->state->color_mgmt_changed) +vop_crtc_gamma_set (vop, crtc, old_crtc_state); + } +} +static void vop_crtc_atomic_flush (struct drm_crtc *crtc,struct drm_crtc_state *old_crtc_state) {@@ -1245,6 +1342,7 @@ static void vop_crtc_atomic_flush(struct drm_crtc *crtc,static const struct drm_crtc_helper_funcs vop_crtc_helper_funcs = { .mode_fixup = vop_crtc_mode_fixup, + .atomic_check = vop_crtc_atomic_check, .atomic_flush = vop_crtc_atomic_flush, .atomic_enable = vop_crtc_atomic_enable, .atomic_disable = vop_crtc_atomic_disable,@@ -1363,6 +1461,7 @@ static const struct drm_crtc_funcs vop_crtc_funcs = { .disable_vblank = vop_crtc_disable_vblank, .set_crc_source = vop_crtc_set_crc_source, .verify_crc_source = vop_crtc_verify_crc_source, + .gamma_set = drm_atomic_helper_legacy_gamma_set, };static void vop_fb_unref_worker (struct drm_flip_work *work,void *val)@@ -1520,6 +1619,10 @@ static int vop_create_crtc(struct vop *vop) goto err_cleanup_planes;drm_crtc_helper_add (crtc, &vop_crtc_helper_funcs); +if (vop->lut_regs) { +drm_mode_crtc_set_gamma_size (crtc, vop_data->lut_size); +drm_crtc_enable_color_mgmt (crtc,0 , false, vop_data->lut_size); + }/* * Create drm_planes for overlay windows with possible_crtcs restricted @@ -1825,6 +1928,17 @@ static int vop_bind(struct device *dev, struct device *master, void *data)if (IS_ERR (vop->regs))return PTR_ERR (vop->regs); + res =platform_get_resource (pdev,IORESOURCE_MEM ,1 ); +if (res) { +if (!vop_data->lut_size) { +DRM_DEV_ERROR (dev,"no gamma LUT size defined\n" ); +return -EINVAL ; + } + vop->lut_regs =devm_ioremap_resource (dev, res); +if (IS_ERR (vop->lut_regs)) +return PTR_ERR (vop->lut_regs); + } + vop->regsbak =devm_kzalloc (dev, vop->len,GFP_KERNEL );if (!vop->regsbak)return -ENOMEM ;drivers/gpu/drm/rockchip/rockchip_drm_vop.h +7
-0@@ -7,6 +7,8 @@#ifndef _ROCKCHIP_DRM_VOP_H#define _ROCKCHIP_DRM_VOP_H +#include <drm/drm_atomic.h> +/* * major: IP major version, used for IP structure * minor: big feature change under same structure @@ -67,6 +69,7 @@ struct vop_common {struct vop_reg dither_down_mode;struct vop_reg dither_down_en;struct vop_reg dither_up; +struct vop_reg dsp_lut_en;struct vop_reg gate_en;struct vop_reg mmu_en;struct vop_reg out_mode;@@ -170,6 +173,7 @@ struct vop_data {const struct vop_win_yuv2yuv_data *win_yuv2yuv;const struct vop_win_data *win;unsigned int win_size; +unsigned int lut_size;#define VOP_FEATURE_OUTPUT_RGB10 BIT(0)#define VOP_FEATURE_INTERNAL_RGB BIT(1)@@ -373,4 +377,7 @@ static inline int scl_vop_cal_lb_mode(int width, bool is_yuv) }extern const struct component_ops vop_component_ops; + +void rockchip_drm_vop_gamma_set (struct drm_atomic_state *state); +#endif /* _ROCKCHIP_DRM_VOP_H */ drivers/gpu/drm/rockchip/rockchip_vop_reg.c +2
-0@@ -598,6 +598,7 @@ static const struct vop_common rk3288_common = { .dither_down_en = VOP_REG(RK3288_DSP_CTRL1, 0x1, 2), .pre_dither_down = VOP_REG(RK3288_DSP_CTRL1, 0x1, 1), .dither_up = VOP_REG(RK3288_DSP_CTRL1, 0x1, 6), + .dsp_lut_en = VOP_REG(RK3288_DSP_CTRL1, 0x1, 0), .data_blank = VOP_REG(RK3288_DSP_CTRL0, 0x1, 19), .dsp_blank = VOP_REG(RK3288_DSP_CTRL0, 0x3, 18), .out_mode = VOP_REG(RK3288_DSP_CTRL0, 0xf, 0),@@ -646,6 +647,7 @@ static const struct vop_data rk3288_vop = { .output = &rk3288_output, .win = rk3288_vop_win_data, .win_size = ARRAY_SIZE(rk3288_vop_win_data), + .lut_size = 1024, };static const int rk3368_vop_intrs[] = {Signature · 1 line
2.22.0
Signature · 5 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v3 3/5] drm/rockchip: Add optional support for CRTC gamma LUT
Hello Sean,
Quoted 61 lines · Sean Paul
Quoted 60 lines · Ezequiel Garcia
Add an optional CRTC gamma LUT support, and enable it on RK3288. This is currently enabled via a separate address resource, which needs to be specified in the devicetree.
The address resource is required because on some SoCs, such as RK3288, the LUT address is after the MMU address, and the latter is supported by a different driver. This prevents the DRM driver from requesting an entire register space.
The current implementation works for RGB 10-bit tables, as that is what seems to work on RK3288.
Signed-off-by Ezequiel Garcia <[email protected]>Reviewed-by Douglas Anderson <[email protected]>Reviewed-by Jacopo Mondi <[email protected]>---
Changes from v2:
* None.Changes from v1:
* drop explicit linear LUT after finding a proper way to disable gamma correction.
* avoid setting gamma is the CRTC is not active.
* s/int/unsigned int as suggested by Jacopo.
* only enable color management and set gamma size if gamma LUT is supported, suggested by Doug.
* drop the reg-names usage, and instead just use indexed reg specifiers, suggested by Doug.Changes from RFC:
* Request (an optional) address resource for the LUT.
* Drop support for RK3399, which doesn't seem to work out of the box and needs more research.
* Support pass-thru setting whenGAMMA_LUTis NULL.
* Add a check for the gamma size, as suggested by Ilia.
* Move gamma setting toatomic_commit_tail, as pointed out by Jacopo/Laurent, is the correct way.--- drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 3 + drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 114 ++++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 7 ++ drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 2 + 4 files changed, 126 insertions(+)drivers/gpu/drm/rockchip/rockchip_drm_fb.c +1
-0@@ -17,6 +17,7 @@#include "rockchip_drm_drv.h" #include "rockchip_drm_fb.h" #include "rockchip_drm_gem.h" +#include "rockchip_drm_vop.h" static const struct drm_framebuffer_funcs rockchip_drm_fb_funcs = { .destroy = drm_gem_fb_destroy,@@ -112,6 +113,8 @@ rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state)drm_atomic_helper_commit_modeset_disables (dev, old_state);
Quoted 9 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state) · line 117
Quoted 2 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +2
-0+rockchip_drm_vop_gamma_set (old_state); +Instead of duplicating the
commit_tailhelper, could you just implement .atomic_begin() and call this from there? I think the only hitch is if you need this to be completed beforecrtc->atomic_enable(), at which point you might need to call it fromvop_crtc_atomic_enable()and then detect that inatomic_begin()
I think moving this to .atomic_begin might be enough. Let me send a new series and we can see how that goes.
Thanks for reviewing,
Ezequiel
Re: [PATCH v3 3/5] drm/rockchip: Add optional support for CRTC gamma LUT
Quoted 74 lines · Ezequiel Garcia
Hello Sean,
Quoted 61 lines · Sean Paul
Quoted 60 lines · Ezequiel Garcia
Add an optional CRTC gamma LUT support, and enable it on RK3288. This is currently enabled via a separate address resource, which needs to be specified in the devicetree.
The address resource is required because on some SoCs, such as RK3288, the LUT address is after the MMU address, and the latter is supported by a different driver. This prevents the DRM driver from requesting an entire register space.
The current implementation works for RGB 10-bit tables, as that is what seems to work on RK3288.
Signed-off-by Ezequiel Garcia <[email protected]>Reviewed-by Douglas Anderson <[email protected]>Reviewed-by Jacopo Mondi <[email protected]>---
Changes from v2:
* None.Changes from v1:
* drop explicit linear LUT after finding a proper way to disable gamma correction.
* avoid setting gamma is the CRTC is not active.
* s/int/unsigned int as suggested by Jacopo.
* only enable color management and set gamma size if gamma LUT is supported, suggested by Doug.
* drop the reg-names usage, and instead just use indexed reg specifiers, suggested by Doug.Changes from RFC:
* Request (an optional) address resource for the LUT.
* Drop support for RK3399, which doesn't seem to work out of the box and needs more research.
* Support pass-thru setting whenGAMMA_LUTis NULL.
* Add a check for the gamma size, as suggested by Ilia.
* Move gamma setting toatomic_commit_tail, as pointed out by Jacopo/Laurent, is the correct way.--- drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 3 + drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 114 ++++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 7 ++ drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 2 + 4 files changed, 126 insertions(+)drivers/gpu/drm/rockchip/rockchip_drm_fb.c +1
-0@@ -17,6 +17,7 @@#include "rockchip_drm_drv.h" #include "rockchip_drm_fb.h" #include "rockchip_drm_gem.h" +#include "rockchip_drm_vop.h" static const struct drm_framebuffer_funcs rockchip_drm_fb_funcs = { .destroy = drm_gem_fb_destroy,@@ -112,6 +113,8 @@ rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state)drm_atomic_helper_commit_modeset_disables (dev, old_state);Quoted 9 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state) · line 117
Quoted 2 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +2
-0+rockchip_drm_vop_gamma_set (old_state); +Instead of duplicating the
commit_tailhelper, could you just implement .atomic_begin() and call this from there? I think the only hitch is if you need this to be completed beforecrtc->atomic_enable(), at which point you might need to call it fromvop_crtc_atomic_enable()and then detect that inatomic_begin()
Quoted 3 lines
I think moving this to .atomic_begin might be enough. Let me send a new series and we can see how that goes.
Oh, before going forward, pleaste note that the first iteration of this patch (as noted in the changelog) was applying the gamma lut on .atomic_flush. However, Laurent and Jacopo pointed out that it might add some tearing to do so, and that's why it was moved to commit_tail.
I have to admit I'm not too sure about the difference between applying this gamma LUT on atomic_begin or atomic_flush, perhaps you can clarify that?
Thanks!
Ezequiel
Re: [PATCH v3 3/5] drm/rockchip: Add optional support for CRTC gamma LUT
Quoted 85 lines · Ezequiel Garcia
Quoted 74 lines · Ezequiel Garcia
Hello Sean,
Quoted 61 lines · Sean Paul
Quoted 60 lines · Ezequiel Garcia
Add an optional CRTC gamma LUT support, and enable it on RK3288. This is currently enabled via a separate address resource, which needs to be specified in the devicetree.
The address resource is required because on some SoCs, such as RK3288, the LUT address is after the MMU address, and the latter is supported by a different driver. This prevents the DRM driver from requesting an entire register space.
The current implementation works for RGB 10-bit tables, as that is what seems to work on RK3288.
Signed-off-by Ezequiel Garcia <[email protected]>Reviewed-by Douglas Anderson <[email protected]>Reviewed-by Jacopo Mondi <[email protected]>---
Changes from v2:
* None.Changes from v1:
* drop explicit linear LUT after finding a proper way to disable gamma correction.
* avoid setting gamma is the CRTC is not active.
* s/int/unsigned int as suggested by Jacopo.
* only enable color management and set gamma size if gamma LUT is supported, suggested by Doug.
* drop the reg-names usage, and instead just use indexed reg specifiers, suggested by Doug.Changes from RFC:
* Request (an optional) address resource for the LUT.
* Drop support for RK3399, which doesn't seem to work out of the box and needs more research.
* Support pass-thru setting whenGAMMA_LUTis NULL.
* Add a check for the gamma size, as suggested by Ilia.
* Move gamma setting toatomic_commit_tail, as pointed out by Jacopo/Laurent, is the correct way.--- drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 3 + drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 114 ++++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 7 ++ drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 2 + 4 files changed, 126 insertions(+)drivers/gpu/drm/rockchip/rockchip_drm_fb.c +1
-0@@ -17,6 +17,7 @@#include "rockchip_drm_drv.h" #include "rockchip_drm_fb.h" #include "rockchip_drm_gem.h" +#include "rockchip_drm_vop.h" static const struct drm_framebuffer_funcs rockchip_drm_fb_funcs = { .destroy = drm_gem_fb_destroy,@@ -112,6 +113,8 @@ rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state)drm_atomic_helper_commit_modeset_disables (dev, old_state);Quoted 9 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state) · line 117
Quoted 2 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +2
-0+rockchip_drm_vop_gamma_set (old_state); +Instead of duplicating the
commit_tailhelper, could you just implement .atomic_begin() and call this from there? I think the only hitch is if you need this to be completed beforecrtc->atomic_enable(), at which point you might need to call it fromvop_crtc_atomic_enable()and then detect that inatomic_begin()Quoted 3 lines
I think moving this to .atomic_begin might be enough. Let me send a new series and we can see how that goes.
Oh, before going forward, pleaste note that the first iteration of this patch (as noted in the changelog) was applying the gamma lut on .atomic_flush. However, Laurent and Jacopo pointed out that it might add some tearing to do so, and that's why it was moved to
commit_tail.
Quoted 3 lines
I have to admit I'm not too sure about the difference between applying this gamma LUT on
atomic_beginoratomic_flush, perhaps you can clarify that?
The only difference between what you have now and calling it in atomic_begin is that as you have it now, it's set before crtc->atomic_enable() is called. I think in order to address Ville's concerns on the other patch, you'll need to set it the lut in .atomic_enable() anyways, so here's what I would suggest:
- Set the LUT in .atomic_enable() wherever it makes sense (you have it at the start now)
- Add an .atomic_begin() implementation and check state->color_mgmt_changed and state->active_changed. color_mgmt_changed && !active_changed, set the lut
- Remove patches 1 & 5
...I think :-)
Sean
Quoted 5 lines
Thanks!
Ezequiel
Signature · 1 line
Sean Paul, Software Engineer, Google / Chromium OS
Re: [PATCH v3 3/5] drm/rockchip: Add optional support for CRTC gamma LUT
Quoted 95 lines · Sean Paul
Quoted 85 lines · Ezequiel Garcia
Quoted 74 lines · Ezequiel Garcia
Hello Sean,
Quoted 61 lines · Sean Paul
Quoted 60 lines · Ezequiel Garcia
Add an optional CRTC gamma LUT support, and enable it on RK3288. This is currently enabled via a separate address resource, which needs to be specified in the devicetree.
The address resource is required because on some SoCs, such as RK3288, the LUT address is after the MMU address, and the latter is supported by a different driver. This prevents the DRM driver from requesting an entire register space.
The current implementation works for RGB 10-bit tables, as that is what seems to work on RK3288.
Signed-off-by Ezequiel Garcia <[email protected]>Reviewed-by Douglas Anderson <[email protected]>Reviewed-by Jacopo Mondi <[email protected]>---
Changes from v2:
* None.Changes from v1:
* drop explicit linear LUT after finding a proper way to disable gamma correction.
* avoid setting gamma is the CRTC is not active.
* s/int/unsigned int as suggested by Jacopo.
* only enable color management and set gamma size if gamma LUT is supported, suggested by Doug.
* drop the reg-names usage, and instead just use indexed reg specifiers, suggested by Doug.Changes from RFC:
* Request (an optional) address resource for the LUT.
* Drop support for RK3399, which doesn't seem to work out of the box and needs more research.
* Support pass-thru setting whenGAMMA_LUTis NULL.
* Add a check for the gamma size, as suggested by Ilia.
* Move gamma setting toatomic_commit_tail, as pointed out by Jacopo/Laurent, is the correct way.--- drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 3 + drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 114 ++++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 7 ++ drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 2 + 4 files changed, 126 insertions(+)drivers/gpu/drm/rockchip/rockchip_drm_fb.c +1
-0@@ -17,6 +17,7 @@#include "rockchip_drm_drv.h" #include "rockchip_drm_fb.h" #include "rockchip_drm_gem.h" +#include "rockchip_drm_vop.h" static const struct drm_framebuffer_funcs rockchip_drm_fb_funcs = { .destroy = drm_gem_fb_destroy,@@ -112,6 +113,8 @@ rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state)drm_atomic_helper_commit_modeset_disables (dev, old_state);Quoted 9 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c rockchip_atomic_helper_commit_tail_rpm(struct drm_atomic_state *old_state) · line 117
Quoted 2 lines
drivers/gpu/drm/rockchip/rockchip_drm_fb.c +2
-0+rockchip_drm_vop_gamma_set (old_state); +Instead of duplicating the
commit_tailhelper, could you just implement .atomic_begin() and call this from there? I think the only hitch is if you need this to be completed beforecrtc->atomic_enable(), at which point you might need to call it fromvop_crtc_atomic_enable()and then detect that inatomic_begin()Quoted 3 lines
I think moving this to .atomic_begin might be enough. Let me send a new series and we can see how that goes.
Oh, before going forward, pleaste note that the first iteration of this patch (as noted in the changelog) was applying the gamma lut on .atomic_flush. However, Laurent and Jacopo pointed out that it might add some tearing to do so, and that's why it was moved to
commit_tail.Quoted 3 lines
I have to admit I'm not too sure about the difference between applying this gamma LUT on
atomic_beginoratomic_flush, perhaps you can clarify that?The only difference between what you have now and calling it in
atomic_beginis that as you have it now, it's set beforecrtc->atomic_enable()is called. I think in order to address Ville's concerns on the other patch, you'll need to set it the lut in .atomic_enable() anyways, so here's what I would suggest:
Quoted 8 lines
- Set the LUT in .atomic_enable() wherever it makes sense (you have it at the start now)
- Add an .atomic_begin() implementation and checkstate->color_mgmt_changedandstate->active_changed.color_mgmt_changed&& !active_changed, set the lut
- Remove patches 1 & 5...I think :-)
OK, that helped! Patches in 3...2...
Regards,
Ezequiel
PATCH v3 4/5 ARM: dts: rockchip: Add RK3288 VOP gamma LUT address
In reply to Ezequiel GarciaRK3288 SoC VOPs have optional support Gamma LUT setting, which requires specifying the Gamma LUT address in the devicetree.
---
Changes from v2:
* None.
Changes from v1:
* Drop reg-names, as suggested by Doug.
--- arch/arm/boot/dts/rk3288.dtsi | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm/boot/dts/rk3288.dtsi +2 -2
@@ -1023,7 +1023,7 @@ vopb: vop@ff930000 { compatible = "rockchip,rk3288-vop"; - reg = <0x0 0xff930000 0x0 0x19c>; + reg = <0x0 0xff930000 0x0 0x19c>, <0x0 0xff931000 0x0 0x1000>; interrupts = <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>; clocks = <&cru ACLK_VOP0>, <&cru DCLK_VOP0>, <&cru HCLK_VOP0>; clock-names = "aclk_vop", "dclk_vop", "hclk_vop";
@@ -1073,7 +1073,7 @@ vopl: vop@ff940000 { compatible = "rockchip,rk3288-vop"; - reg = <0x0 0xff940000 0x0 0x19c>; + reg = <0x0 0xff940000 0x0 0x19c>, <0x0 0xff941000 0x0 0x1000>; interrupts = <GIC_SPI 16 IRQ_TYPE_LEVEL_HIGH>; clocks = <&cru ACLK_VOP1>, <&cru DCLK_VOP1>, <&cru HCLK_VOP1>; clock-names = "aclk_vop", "dclk_vop", "hclk_vop";
Signature · 1 line
2.22.0
PATCH v3 5/5 RFC: drm/atomic-helper: Reapply color transformation after resume
In reply to Ezequiel GarciaSome platforms are not able to maintain the color transformation state after a system suspend/resume cycle.
Set the colog_mgmt_changed flag so that CMM on the CRTCs in the suspend state are reapplied after system resume.
---
This is an RFC, and it's mostly based on Jacopo Mondi's work https://lkml.org/lkml/2019/9/6/498.
Changes from v2:
* New patch.
--- drivers/gpu/drm/drm_atomic_helper.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +12 -0
@@ -3234,8 +3234,20 @@ int drm_atomic_helper_resume(struct drm_device *dev, struct drm_atomic_state *state) {struct drm_modeset_acquire_ctx ctx; +struct drm_crtc_state *crtc_state; +struct drm_crtc *crtc; +unsigned int i;int err; +for_each_new_crtc_in_state (state, crtc, crtc_state, i) { +/* +* Force re-enablement of CMM after system resume if any +* of the DRM color transformation properties was set in +* the state saved at system suspend time. +*/ +if (crtc_state->gamma_lut) + crtc_state->color_mgmt_changed = true; + }drm_mode_config_reset (dev);DRM_MODESET_LOCK_ALL_BEGIN (dev, ctx,0 , err);
Signature · 1 line
2.22.0
Re: [PATCH v3 5/5] RFC: drm/atomic-helper: Reapply color transformation after resume
Hi Ezequiel,
Quoted 4 lines · Ezequiel Garcia
Some platforms are not able to maintain the color transformation state after a system suspend/resume cycle.
Set the
colog_mgmt_changedflag so that CMM on the CRTCs in
CMM is the name of the Renesas unit for color enanchement. It should not be used here as this will apply to all platforms.
Quoted 22 lines
the suspend state are reapplied after system resume.
Signed-off-by Ezequiel Garcia <[email protected]>---
This is an RFC, and it's mostly based on Jacopo Mondi's work https://lkml.org/lkml/2019/9/6/498.Changes from v2:
* New patch.--- drivers/gpu/drm/drm_atomic_helper.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)drivers/gpu/drm/drm_atomic_helper.c +2
-0@@ -3234,8 +3234,20 @@ int drm_atomic_helper_resume(struct drm_device *dev, struct drm_atomic_state *state) {struct drm_modeset_acquire_ctx ctx; +struct drm_crtc_state *crtc_state; +struct drm_crtc *crtc;
drivers/gpu/drm/drm_atomic_helper.c int drm_atomic_helper_resume(struct drm_device *dev, · line 3248
Quoted 10 lines
drivers/gpu/drm/drm_atomic_helper.c +8
-0+unsigned int i;int err; +for_each_new_crtc_in_state (state, crtc, crtc_state, i) { +/* +* Force re-enablement of CMM after system resume if any +* of the DRM color transformation properties was set in +* the state saved at system suspend time. +*/ +if (crtc_state->gamma_lut)
Please note that in my original patch I only took gamma_lut into account as that's what our CMM supports at the moment, but you should here consider the degamma_lut and cmt flags in the crtc_state.
Quoted 8 lines
drivers/gpu/drm/drm_atomic_helper.c +2
-1+ crtc_state->color_mgmt_changed = true; + }drm_mode_config_reset (dev);DRM_MODESET_LOCK_ALL_BEGIN (dev, ctx,0 , err); --2.22.0
Re: [PATCH v3 5/5] RFC: drm/atomic-helper: Reapply color transformation after resume
In reply to Ezequiel GarciaQuoted 26 lines · Ezequiel Garcia
Some platforms are not able to maintain the color transformation state after a system suspend/resume cycle.
Set the
colog_mgmt_changedflag so that CMM on the CRTCs in the suspend state are reapplied after system resume.Signed-off-by Ezequiel Garcia <[email protected]>---
This is an RFC, and it's mostly based on Jacopo Mondi's work https://lkml.org/lkml/2019/9/6/498.Changes from v2:
* New patch.--- drivers/gpu/drm/drm_atomic_helper.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)drivers/gpu/drm/drm_atomic_helper.c +2
-0@@ -3234,8 +3234,20 @@ int drm_atomic_helper_resume(struct drm_device *dev, struct drm_atomic_state *state) {struct drm_modeset_acquire_ctx ctx; +struct drm_crtc_state *crtc_state; +struct drm_crtc *crtc;
drivers/gpu/drm/drm_atomic_helper.c int drm_atomic_helper_resume(struct drm_device *dev, · line 3248
Quoted 10 lines
drivers/gpu/drm/drm_atomic_helper.c +8
-0+unsigned int i;int err; +for_each_new_crtc_in_state (state, crtc, crtc_state, i) { +/* +* Force re-enablement of CMM after system resume if any +* of the DRM color transformation properties was set in +* the state saved at system suspend time. +*/ +if (crtc_state->gamma_lut)
You say "any" but you check the one?
drivers/gpu/drm/drm_atomic_helper.c int drm_atomic_helper_resume(struct drm_device *dev, · line 3249
Quoted 1 line
drivers/gpu/drm/drm_atomic_helper.c +1
-0+ crtc_state->color_mgmt_changed = true;
But I'm not convinced this is the best way to go about it. I would generally expect that you repgrogram everything when doing a full modeset since the state was possibly lost while the crtc was disabled.
Quoted 11 lines
drivers/gpu/drm/drm_atomic_helper.c +1
-0+ }drm_mode_config_reset (dev);DRM_MODESET_LOCK_ALL_BEGIN (dev, ctx,0 , err);Signature · 6 lines
2.22.0 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 2 lines
Ville Syrjälä Intel