Skip to contentSkip to conversation

drm/rockchip: vop: clear DMA stop bit upon vblank on RK3066

PATCH Superseded Val Packett 2 participants 4 messages May 27, 2024 – Jun 8, 2024

Subsystems (5)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/2 drm/rockchip: vop: clear DMA stop bit upon vblank on RK3066

The RK3066 VOP sets a dma_stop bit when it's done scanning out a frame and needs the driver to acknowledge that by clearing the bit.

So unless we clear it "between" frames, the RGB output only shows noise instead of the picture. vblank seems to be the most appropriate place to do it, since it indicates exactly that: that the hardware is done with the frame.

This seems to be a redundant synchronization mechanism that was removed in later iterations of the VOP hardware block.

Fixes f4a6de8 ("drm: rockchip: vop: add rk3066 vop definitions")
---
 drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 6 ++++++
 drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 1 +
 drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 1 +
 3 files changed, 8 insertions(+)
drivers/gpu/drm/rockchip/rockchip_drm_vop.c +6 -0
index a13473b2d..2731fe2b2 100644
@@ -1766,6 +1766,12 @@ static void vop_handle_vblank(struct vop *vop)
 	}
 	spin_unlock(&drm->event_lock);
 
+	if (VOP_HAS_REG(vop, common, dma_stop)) {
+		spin_lock(&vop->reg_lock);
+		VOP_REG_SET(vop, common, dma_stop, 0);
+		spin_unlock(&vop->reg_lock);
+	}
+
 	if (test_and_clear_bit(VOP_PENDING_FB_UNREF, &vop->pending))
 		drm_flip_work_commit(&vop->fb_unref_work, system_unbound_wq);
 }
drivers/gpu/drm/rockchip/rockchip_drm_vop.h +1 -0
index b33e5bdc2..0cf512cc1 100644
@@ -122,6 +122,7 @@ struct vop_common {
 	struct vop_reg lut_buffer_index;
 	struct vop_reg gate_en;
 	struct vop_reg mmu_en;
+	struct vop_reg dma_stop;
 	struct vop_reg out_mode;
 	struct vop_reg standby;
 };
drivers/gpu/drm/rockchip/rockchip_vop_reg.c +1 -0
index b9ee02061..9bcb40a64 100644
@@ -466,6 +466,7 @@ static const struct vop_output rk3066_output = {
 };
 
 static const struct vop_common rk3066_common = {
+	.dma_stop = VOP_REG(RK3066_SYS_CTRL0, 0x1, 0),
 	.standby = VOP_REG(RK3066_SYS_CTRL0, 0x1, 1),
 	.out_mode = VOP_REG(RK3066_DSP_CTRL0, 0xf, 0),
 	.cfg_done = VOP_REG(RK3066_REG_CFG_DONE, 0x1, 0),
Signature · 1 line
2.45.1
#1 lore raw

PATCH v3 2/2 drm/rockchip: vop: enable VOP_FEATURE_INTERNAL_RGB on RK3066

The RK3066 does have RGB display output, so it should be marked as such.

Fixes f4a6de8 ("drm: rockchip: vop: add rk3066 vop definitions")
---
 drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 1 +
 1 file changed, 1 insertion(+)
drivers/gpu/drm/rockchip/rockchip_vop_reg.c +1 -0
index 9bcb40a64..e2c6ba26f 100644
@@ -515,6 +515,7 @@ static const struct vop_data rk3066_vop = {
 	.output = &rk3066_output,
 	.win = rk3066_vop_win_data,
 	.win_size = ARRAY_SIZE(rk3066_vop_win_data),
+	.feature = VOP_FEATURE_INTERNAL_RGB,
 	.max_output = { 1920, 1080 },
 };
 
Signature · 1 line
2.45.1
#2 lore raw
In reply to Val Packett
Quoted 25 lines · Val Packett

The RK3066 VOP sets a dma_stop bit when it's done scanning out a frame and needs the driver to acknowledge that by clearing the bit.

So unless we clear it "between" frames, the RGB output only shows noise instead of the picture. vblank seems to be the most appropriate place to do it, since it indicates exactly that: that the hardware is done with the frame.

This seems to be a redundant synchronization mechanism that was removed in later iterations of the VOP hardware block.

Fixes f4a6de8 ("drm: rockchip: vop: add rk3066 vop definitions")
---
 drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 6 ++++++
 drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 1 +
 drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 1 +
 3 files changed, 8 insertions(+)
drivers/gpu/drm/rockchip/rockchip_drm_vop.c +0 -0
index a13473b2d..2731fe2b2 100644

drivers/gpu/drm/rockchip/rockchip_drm_vop.c static void vop_handle_vblank(struct vop *vop) · line 1774

Quoted 9 lines
drivers/gpu/drm/rockchip/rockchip_drm_vop.c +6 -0
@@ -1766,6 +1766,12 @@ static void vop_handle_vblank(struct vop *vop)
 	}
 	spin_unlock(&drm->event_lock);
 
+	if (VOP_HAS_REG(vop, common, dma_stop)) {
+		spin_lock(&vop->reg_lock);
+		VOP_REG_SET(vop, common, dma_stop, 0);
+		spin_unlock(&vop->reg_lock);
+	}
+

Oops… so doing it here actually causes deadlocks, unless we also change all other reg_lock usages to be spin_lock_irq/spin_unlock_irq.

Not sure if doing that or going back to v1 would be better.

maintainer
#3 lore raw

Am Sonntag, 2. Juni 2024, 05:35:36 CEST schrieb Val Packett:

Quoted 45 lines
Quoted 29 lines · Val Packett

The RK3066 VOP sets a dma_stop bit when it's done scanning out a frame and needs the driver to acknowledge that by clearing the bit.

So unless we clear it "between" frames, the RGB output only shows noise
instead of the picture. vblank seems to be the most appropriate place to
do it, since it indicates exactly that: that the hardware is done with the frame.

This seems to be a redundant synchronization mechanism that was removed
in later iterations of the VOP hardware block.

Fixes f4a6de8 ("drm: rockchip: vop: add rk3066 vop definitions")
---
 drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 6 ++++++
 drivers/gpu/drm/rockchip/rockchip_drm_vop.h | 1 +
 drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 1 +
 3 files changed, 8 insertions(+)
drivers/gpu/drm/rockchip/rockchip_drm_vop.c +0 -0
index a13473b2d..2731fe2b2 100644

drivers/gpu/drm/rockchip/rockchip_drm_vop.c static void vop_handle_vblank(struct vop *vop) · line 1774

Quoted 9 lines
drivers/gpu/drm/rockchip/rockchip_drm_vop.c +6 -0
@@ -1766,6 +1766,12 @@ static void vop_handle_vblank(struct vop *vop)
 	}
 	spin_unlock(&drm->event_lock);
 
+	if (VOP_HAS_REG(vop, common, dma_stop)) {
+		spin_lock(&vop->reg_lock);
+		VOP_REG_SET(vop, common, dma_stop, 0);
+		spin_unlock(&vop->reg_lock);
+	}
+

Oops… so doing it here actually causes deadlocks, unless we also change all other reg_lock usages to be spin_lock_irq/spin_unlock_irq.

Quoted 1 line

Not sure if doing that or going back to v1 would be better.

then please go back to v1 (as v4) :-) .

I.e. regular spinlock vs. spin_lock_irq will have performance implications and this "feature" is a one-time only thing used only on a more than 10 year old soc, so it really must not affect stuff coming after it.

Heiko