* Claude review: drm: verisilicon: add hardware cursor support
2026-03-09 8:53 [PATCH -next 0/2] " Icenowy Zheng
@ 2026-03-10 2:36 ` Claude Code Review Bot
0 siblings, 0 replies; 8+ messages in thread
From: Claude Code Review Bot @ 2026-03-10 2:36 UTC (permalink / raw)
To: dri-devel-reviews
Overall Series Review
Subject: drm: verisilicon: add hardware cursor support
Author: Icenowy Zheng <zhengxingda@iscas.ac.cn>
Patches: 3
Reviewed: 2026-03-10T12:36:01.366419
---
This 2-patch series adds hardware cursor plane support to the Verisilicon display controller driver. Patch 1 extends the HWDB with a max cursor size field, and Patch 2 implements the cursor plane itself. The structure follows the existing primary plane pattern in the driver well.
However, there are **two bugs** in the register programming of the cursor plane that will cause incorrect behavior for negative cursor coordinates, plus off-by-one errors in the coordinate range checks. There are also minor style and cosmetic issues.
---
Generated by Claude Code Patch Reviewer
^ permalink raw reply [flat|nested] 8+ messages in thread
* Claude review: drm: verisilicon: add hardware cursor support
2026-03-24 6:02 [PATCH -next v2 0/2] " Icenowy Zheng
@ 2026-03-24 21:03 ` Claude Code Review Bot
0 siblings, 0 replies; 8+ messages in thread
From: Claude Code Review Bot @ 2026-03-24 21:03 UTC (permalink / raw)
To: dri-devel-reviews
Overall Series Review
Subject: drm: verisilicon: add hardware cursor support
Author: Icenowy Zheng <zhengxingda@iscas.ac.cn>
Patches: 3
Reviewed: 2026-03-25T07:03:50.855249
---
This is a clean, well-structured 2-patch series adding hardware cursor plane support to the Verisilicon DRM driver. The approach — adding HWDB metadata first, then implementing the cursor plane — is logical. The code is reasonably well written and follows existing driver conventions. There are several issues to address, mostly around missing `FIELD_PREP` usage for register field values and a missing `mode_config.cursor_width/height` setup.
---
Generated by Claude Code Patch Reviewer
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH v3 0/2] drm: verisilicon: add hardware cursor support
@ 2026-05-06 17:56 Icenowy Zheng
2026-05-06 17:56 ` [PATCH v3 1/2] drm: verisilicon: add max cursor size to HWDB Icenowy Zheng
` (2 more replies)
0 siblings, 3 replies; 8+ messages in thread
From: Icenowy Zheng @ 2026-05-06 17:56 UTC (permalink / raw)
To: Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann, David Airlie,
Simona Vetter
Cc: linux-kernel, dri-devel, Han Gao, Icenowy Zheng, Icenowy Zheng
Verisilicon display controllers have the hardware cursor feature.
The hardware cursor layer is a simplistic and restrictive layer as the
topmost layer of the display output. It can only support ARGB8888 images
and some certain square power-of-two sizes (32x32 is supported on all
DCs).
This patchset adds support for the hardware cursor as a cursor plane.
Icenowy Zheng (2):
drm: verisilicon: add max cursor size to HWDB
drm: verisilicon: add support for cursor planes
drivers/gpu/drm/verisilicon/Makefile | 3 +-
drivers/gpu/drm/verisilicon/vs_crtc.c | 11 +-
drivers/gpu/drm/verisilicon/vs_cursor_plane.c | 272 ++++++++++++++++++
.../drm/verisilicon/vs_cursor_plane_regs.h | 44 +++
drivers/gpu/drm/verisilicon/vs_hwdb.c | 4 +
drivers/gpu/drm/verisilicon/vs_hwdb.h | 5 +
drivers/gpu/drm/verisilicon/vs_plane.h | 1 +
7 files changed, 337 insertions(+), 3 deletions(-)
create mode 100644 drivers/gpu/drm/verisilicon/vs_cursor_plane.c
create mode 100644 drivers/gpu/drm/verisilicon/vs_cursor_plane_regs.h
--
2.52.0
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH v3 1/2] drm: verisilicon: add max cursor size to HWDB
2026-05-06 17:56 [PATCH v3 0/2] drm: verisilicon: add hardware cursor support Icenowy Zheng
@ 2026-05-06 17:56 ` Icenowy Zheng
2026-05-07 3:14 ` Claude review: " Claude Code Review Bot
2026-05-06 17:56 ` [PATCH v3 2/2] drm: verisilicon: add support for cursor planes Icenowy Zheng
2026-05-07 3:14 ` Claude review: drm: verisilicon: add hardware cursor support Claude Code Review Bot
2 siblings, 1 reply; 8+ messages in thread
From: Icenowy Zheng @ 2026-05-06 17:56 UTC (permalink / raw)
To: Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann, David Airlie,
Simona Vetter
Cc: linux-kernel, dri-devel, Han Gao, Icenowy Zheng, Icenowy Zheng
Different display controller variants support different maximum cursor
size. All known DC8200 variants support both 32x32 and 64x64, but some
DC8000 variants support either only 32x32 or up to 256x256.
The minimum size is fixed at 32 and only PoT square sizes are supported.
Add the max cursor size field to HWDB and fill all entries with 64.
Signed-off-by: Icenowy Zheng <zhengxingda@iscas.ac.cn>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
---
Changes in v3:
- Add R-b from Thomas.
Changes in v2:
- Added a comment about the cursor being square.
drivers/gpu/drm/verisilicon/vs_hwdb.c | 4 ++++
drivers/gpu/drm/verisilicon/vs_hwdb.h | 5 +++++
2 files changed, 9 insertions(+)
diff --git a/drivers/gpu/drm/verisilicon/vs_hwdb.c b/drivers/gpu/drm/verisilicon/vs_hwdb.c
index 09336af0900ae..2a0f7c59afa3a 100644
--- a/drivers/gpu/drm/verisilicon/vs_hwdb.c
+++ b/drivers/gpu/drm/verisilicon/vs_hwdb.c
@@ -95,6 +95,7 @@ static struct vs_chip_identity vs_chip_identities[] = {
.customer_id = ~0U,
.display_count = 2,
+ .max_cursor_size = 64,
.formats = &vs_formats_no_yuv444,
},
{
@@ -103,6 +104,7 @@ static struct vs_chip_identity vs_chip_identities[] = {
.customer_id = 0x30B,
.display_count = 2,
+ .max_cursor_size = 64,
.formats = &vs_formats_no_yuv444,
},
{
@@ -111,6 +113,7 @@ static struct vs_chip_identity vs_chip_identities[] = {
.customer_id = 0x310,
.display_count = 2,
+ .max_cursor_size = 64,
.formats = &vs_formats_with_yuv444,
},
{
@@ -119,6 +122,7 @@ static struct vs_chip_identity vs_chip_identities[] = {
.customer_id = 0x311,
.display_count = 2,
+ .max_cursor_size = 64,
.formats = &vs_formats_no_yuv444,
},
};
diff --git a/drivers/gpu/drm/verisilicon/vs_hwdb.h b/drivers/gpu/drm/verisilicon/vs_hwdb.h
index 92192e4fa0862..2065ecb730437 100644
--- a/drivers/gpu/drm/verisilicon/vs_hwdb.h
+++ b/drivers/gpu/drm/verisilicon/vs_hwdb.h
@@ -20,6 +20,11 @@ struct vs_chip_identity {
u32 customer_id;
u32 display_count;
+ /*
+ * The hardware only supports square cursor planes, so this field
+ * is both the maximum width and height in pixels.
+ */
+ int32_t max_cursor_size;
const struct vs_formats *formats;
};
--
2.52.0
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH v3 2/2] drm: verisilicon: add support for cursor planes
2026-05-06 17:56 [PATCH v3 0/2] drm: verisilicon: add hardware cursor support Icenowy Zheng
2026-05-06 17:56 ` [PATCH v3 1/2] drm: verisilicon: add max cursor size to HWDB Icenowy Zheng
@ 2026-05-06 17:56 ` Icenowy Zheng
2026-05-07 3:14 ` Claude review: " Claude Code Review Bot
2026-05-07 3:14 ` Claude review: drm: verisilicon: add hardware cursor support Claude Code Review Bot
2 siblings, 1 reply; 8+ messages in thread
From: Icenowy Zheng @ 2026-05-06 17:56 UTC (permalink / raw)
To: Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann, David Airlie,
Simona Vetter
Cc: linux-kernel, dri-devel, Han Gao, Icenowy Zheng, Icenowy Zheng
Verisilicon display controllers support hardware cursors per output
port.
Add support for them as cursor planes.
Signed-off-by: Icenowy Zheng <zhengxingda@iscas.ac.cn>
---
Changes in v3:
- Moved the validity check for max_cursor_size to
vs_cursor_plane_init().
- Removed NULL check for new_plane_state->fb because
drm_atomic_helper_check_plane_state() will set visibility to false in
such case.
Changes in v2:
- Use is_power_of_two instead of switch to validate cursor size.
- Remove unneeded format check.
- Skip the whole validity check if the plane is invisible (previously
it's only skipped when the plane is not bound to CRTC).
- Explicitly set modifiers list (linear only).
drivers/gpu/drm/verisilicon/Makefile | 3 +-
drivers/gpu/drm/verisilicon/vs_crtc.c | 11 +-
drivers/gpu/drm/verisilicon/vs_cursor_plane.c | 272 ++++++++++++++++++
.../drm/verisilicon/vs_cursor_plane_regs.h | 44 +++
drivers/gpu/drm/verisilicon/vs_plane.h | 1 +
5 files changed, 328 insertions(+), 3 deletions(-)
create mode 100644 drivers/gpu/drm/verisilicon/vs_cursor_plane.c
create mode 100644 drivers/gpu/drm/verisilicon/vs_cursor_plane_regs.h
diff --git a/drivers/gpu/drm/verisilicon/Makefile b/drivers/gpu/drm/verisilicon/Makefile
index fd8d805fbcde1..426f4bcaa834d 100644
--- a/drivers/gpu/drm/verisilicon/Makefile
+++ b/drivers/gpu/drm/verisilicon/Makefile
@@ -1,5 +1,6 @@
# SPDX-License-Identifier: GPL-2.0-only
-verisilicon-dc-objs := vs_bridge.o vs_crtc.o vs_dc.o vs_drm.o vs_hwdb.o vs_plane.o vs_primary_plane.o
+verisilicon-dc-objs := vs_bridge.o vs_crtc.o vs_dc.o vs_drm.o vs_hwdb.o \
+ vs_plane.o vs_primary_plane.o vs_cursor_plane.o
obj-$(CONFIG_DRM_VERISILICON_DC) += verisilicon-dc.o
diff --git a/drivers/gpu/drm/verisilicon/vs_crtc.c b/drivers/gpu/drm/verisilicon/vs_crtc.c
index f494017130006..5c9714a3e69a7 100644
--- a/drivers/gpu/drm/verisilicon/vs_crtc.c
+++ b/drivers/gpu/drm/verisilicon/vs_crtc.c
@@ -159,7 +159,7 @@ struct vs_crtc *vs_crtc_init(struct drm_device *drm_dev, struct vs_dc *dc,
unsigned int output)
{
struct vs_crtc *vcrtc;
- struct drm_plane *primary;
+ struct drm_plane *primary, *cursor;
int ret;
vcrtc = drmm_kzalloc(drm_dev, sizeof(*vcrtc), GFP_KERNEL);
@@ -175,9 +175,16 @@ struct vs_crtc *vs_crtc_init(struct drm_device *drm_dev, struct vs_dc *dc,
return ERR_PTR(PTR_ERR(primary));
}
+ /* Create our cursor plane */
+ cursor = vs_cursor_plane_init(drm_dev, dc);
+ if (IS_ERR(cursor)) {
+ drm_err(drm_dev, "Couldn't create the cursor plane\n");
+ return ERR_CAST(cursor);
+ }
+
ret = drmm_crtc_init_with_planes(drm_dev, &vcrtc->base,
primary,
- NULL,
+ cursor,
&vs_crtc_funcs,
NULL);
if (ret) {
diff --git a/drivers/gpu/drm/verisilicon/vs_cursor_plane.c b/drivers/gpu/drm/verisilicon/vs_cursor_plane.c
new file mode 100644
index 0000000000000..acb9854fa67ed
--- /dev/null
+++ b/drivers/gpu/drm/verisilicon/vs_cursor_plane.c
@@ -0,0 +1,272 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2026 Institute of Software, Chinese Academy of Sciences (ISCAS)
+ *
+ * Authors:
+ * Icenowy Zheng <zhengxingda@iscas.ac.cn>
+ */
+
+#include <linux/log2.h>
+#include <linux/regmap.h>
+
+#include <drm/drm_atomic.h>
+#include <drm/drm_atomic_helper.h>
+#include <drm/drm_crtc.h>
+#include <drm/drm_fourcc.h>
+#include <drm/drm_framebuffer.h>
+#include <drm/drm_gem_atomic_helper.h>
+#include <drm/drm_modeset_helper_vtables.h>
+#include <drm/drm_plane.h>
+#include <drm/drm_print.h>
+
+#include "vs_crtc.h"
+#include "vs_plane.h"
+#include "vs_dc.h"
+#include "vs_hwdb.h"
+#include "vs_cursor_plane_regs.h"
+
+#define VSDC_MIN_CURSOR_SIZE 32
+#define VSDC_MAX_CURSOR_SIZE 256
+
+#define VSDC_CURSOR_LOCATION_MAX_POSITIVE BIT_MASK(15)
+#define VSDC_CURSOR_LOCATION_MAX_NEGATIVE BIT_MASK(5)
+
+static bool vs_cursor_plane_check_coord(int32_t coord)
+{
+ if (coord >= 0)
+ return coord <= VSDC_CURSOR_LOCATION_MAX_POSITIVE;
+ else
+ return (-coord) <= VSDC_CURSOR_LOCATION_MAX_NEGATIVE;
+}
+
+static int vs_cursor_plane_atomic_check(struct drm_plane *plane,
+ struct drm_atomic_state *state)
+{
+ struct drm_plane_state *new_plane_state = drm_atomic_get_new_plane_state(state,
+ plane);
+ struct drm_crtc *crtc = new_plane_state->crtc;
+ struct drm_framebuffer *fb = new_plane_state->fb;
+ struct drm_crtc_state *crtc_state = NULL;
+ struct vs_crtc *vcrtc;
+ struct vs_dc *dc;
+ int ret;
+
+ if (crtc)
+ crtc_state = drm_atomic_get_new_crtc_state(state, crtc);
+
+ ret = drm_atomic_helper_check_plane_state(new_plane_state,
+ crtc_state,
+ DRM_PLANE_NO_SCALING,
+ DRM_PLANE_NO_SCALING,
+ true, true);
+ if (ret)
+ return ret;
+
+ if (!new_plane_state->visible)
+ return 0; /* Skip validity check */
+
+ vcrtc = drm_crtc_to_vs_crtc(crtc);
+ dc = vcrtc->dc;
+
+ /* Only certain PoT square sizes is supported. */
+ if (!is_power_of_2(new_plane_state->crtc_w) ||
+ new_plane_state->crtc_w < VSDC_MIN_CURSOR_SIZE ||
+ new_plane_state->crtc_w > dc->identity.max_cursor_size)
+ return -EINVAL;
+
+ if (new_plane_state->crtc_w != new_plane_state->crtc_h)
+ return -EINVAL;
+
+ /* Check if the cursor is inside the register fields' range */
+ if (!vs_cursor_plane_check_coord(new_plane_state->crtc_x) ||
+ !vs_cursor_plane_check_coord(new_plane_state->crtc_y))
+ return -EINVAL;
+
+ /* Extra line padding isn't supported */
+ if (fb->pitches[0] !=
+ drm_format_info_min_pitch(fb->format, 0, new_plane_state->crtc_w))
+ return -EINVAL;
+
+ return 0;
+}
+
+static void vs_cursor_plane_commit(struct vs_dc *dc, unsigned int output)
+{
+ regmap_set_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_COMMIT |
+ VSDC_CURSOR_CONFIG_IMG_UPDATE);
+}
+
+static void vs_cursor_plane_atomic_enable(struct drm_plane *plane,
+ struct drm_atomic_state *atomic_state)
+{
+ struct drm_plane_state *state = drm_atomic_get_new_plane_state(atomic_state,
+ plane);
+ struct drm_crtc *crtc = state->crtc;
+ struct vs_crtc *vcrtc = drm_crtc_to_vs_crtc(crtc);
+ unsigned int output = vcrtc->id;
+ struct vs_dc *dc = vcrtc->dc;
+
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_FMT_MASK,
+ VSDC_CURSOR_CONFIG_FMT_ARGB8888);
+
+ vs_cursor_plane_commit(dc, output);
+}
+
+static void vs_cursor_plane_atomic_disable(struct drm_plane *plane,
+ struct drm_atomic_state *atomic_state)
+{
+ struct drm_plane_state *state = drm_atomic_get_old_plane_state(atomic_state,
+ plane);
+ struct drm_crtc *crtc = state->crtc;
+ struct vs_crtc *vcrtc = drm_crtc_to_vs_crtc(crtc);
+ unsigned int output = vcrtc->id;
+ struct vs_dc *dc = vcrtc->dc;
+
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_FMT_MASK,
+ VSDC_CURSOR_CONFIG_FMT_OFF);
+
+ vs_cursor_plane_commit(dc, output);
+}
+
+static void vs_cursor_plane_atomic_update(struct drm_plane *plane,
+ struct drm_atomic_state *atomic_state)
+{
+ struct drm_plane_state *state = drm_atomic_get_new_plane_state(atomic_state,
+ plane);
+ struct drm_framebuffer *fb = state->fb;
+ struct drm_crtc *crtc = state->crtc;
+ struct vs_dc *dc;
+ struct vs_crtc *vcrtc;
+ unsigned int output;
+ dma_addr_t dma_addr;
+
+ if (!state->visible) {
+ vs_cursor_plane_atomic_disable(plane, atomic_state);
+ return;
+ }
+
+ vcrtc = drm_crtc_to_vs_crtc(crtc);
+ output = vcrtc->id;
+ dc = vcrtc->dc;
+
+ /* Other sizes should be rejected by atomic_check */
+ switch (state->crtc_w) {
+ case 32:
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_SIZE_MASK,
+ VSDC_CURSOR_CONFIG_SIZE_32);
+ break;
+ case 64:
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_SIZE_MASK,
+ VSDC_CURSOR_CONFIG_SIZE_64);
+ break;
+ case 128:
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_SIZE_MASK,
+ VSDC_CURSOR_CONFIG_SIZE_128);
+ break;
+ case 256:
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_SIZE_MASK,
+ VSDC_CURSOR_CONFIG_SIZE_256);
+ break;
+ }
+
+ dma_addr = vs_fb_get_dma_addr(fb, &state->src);
+
+ regmap_write(dc->regs, VSDC_CURSOR_ADDRESS(output),
+ lower_32_bits(dma_addr));
+
+ /*
+ * The X_OFF and Y_OFF fields define which point does the LOCATION
+ * register represent in the cursor image, and LOCATION register
+ * values are unsigned. To for positive left-top coordinates the
+ * offset is set to 0 and the location is set to the coordinate, for
+ * negative coordinates the location is set to 0 and the offset
+ * is set to the opposite number of the coordinate to offset the
+ * cursor image partly off-screen.
+ */
+ if (state->crtc_x >= 0) {
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_X_OFF_MASK, 0);
+ regmap_update_bits(dc->regs, VSDC_CURSOR_LOCATION(output),
+ VSDC_CURSOR_LOCATION_X_MASK,
+ VSDC_CURSOR_LOCATION_X(state->crtc_x));
+ } else {
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_X_OFF_MASK,
+ -state->crtc_x);
+ regmap_update_bits(dc->regs, VSDC_CURSOR_LOCATION(output),
+ VSDC_CURSOR_LOCATION_X_MASK, 0);
+ }
+
+ if (state->crtc_y >= 0) {
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_Y_OFF_MASK, 0);
+ regmap_update_bits(dc->regs, VSDC_CURSOR_LOCATION(output),
+ VSDC_CURSOR_LOCATION_Y_MASK,
+ VSDC_CURSOR_LOCATION_Y(state->crtc_y));
+ } else {
+ regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
+ VSDC_CURSOR_CONFIG_Y_OFF_MASK,
+ -state->crtc_y);
+ regmap_update_bits(dc->regs, VSDC_CURSOR_LOCATION(output),
+ VSDC_CURSOR_LOCATION_Y_MASK, 0);
+ }
+
+ vs_cursor_plane_commit(dc, output);
+}
+
+static const struct drm_plane_helper_funcs vs_cursor_plane_helper_funcs = {
+ .atomic_check = vs_cursor_plane_atomic_check,
+ .atomic_update = vs_cursor_plane_atomic_update,
+ .atomic_enable = vs_cursor_plane_atomic_enable,
+ .atomic_disable = vs_cursor_plane_atomic_disable,
+};
+
+static const struct drm_plane_funcs vs_cursor_plane_funcs = {
+ .atomic_destroy_state = drm_atomic_helper_plane_destroy_state,
+ .atomic_duplicate_state = drm_atomic_helper_plane_duplicate_state,
+ .disable_plane = drm_atomic_helper_disable_plane,
+ .reset = drm_atomic_helper_plane_reset,
+ .update_plane = drm_atomic_helper_update_plane,
+};
+
+static const u32 vs_cursor_plane_formats[] = {
+ DRM_FORMAT_ARGB8888,
+};
+
+static const u64 vs_cursor_plane_modifiers[] = {
+ DRM_FORMAT_MOD_LINEAR,
+ DRM_FORMAT_MOD_INVALID, /* sentinel */
+};
+
+struct drm_plane *vs_cursor_plane_init(struct drm_device *drm_dev,
+ struct vs_dc *dc)
+{
+ int32_t max_cursor_size = dc->identity.max_cursor_size;
+ struct drm_plane *plane;
+
+ if (drm_WARN_ON_ONCE(drm_dev, max_cursor_size < VSDC_MIN_CURSOR_SIZE ||
+ max_cursor_size > VSDC_MAX_CURSOR_SIZE))
+ return ERR_PTR(-EINVAL);
+
+ plane = drmm_universal_plane_alloc(drm_dev, struct drm_plane, dev, 0,
+ &vs_cursor_plane_funcs,
+ vs_cursor_plane_formats,
+ ARRAY_SIZE(vs_cursor_plane_formats),
+ vs_cursor_plane_modifiers,
+ DRM_PLANE_TYPE_CURSOR,
+ NULL);
+
+ if (IS_ERR(plane))
+ return plane;
+
+ drm_plane_helper_add(plane, &vs_cursor_plane_helper_funcs);
+
+ return plane;
+}
diff --git a/drivers/gpu/drm/verisilicon/vs_cursor_plane_regs.h b/drivers/gpu/drm/verisilicon/vs_cursor_plane_regs.h
new file mode 100644
index 0000000000000..99693f2c95b94
--- /dev/null
+++ b/drivers/gpu/drm/verisilicon/vs_cursor_plane_regs.h
@@ -0,0 +1,44 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2025 Icenowy Zheng <uwu@icenowy.me>
+ *
+ * Based on vs_dc_hw.h, which is:
+ * Copyright (C) 2023 VeriSilicon Holdings Co., Ltd.
+ */
+
+#ifndef _VS_CURSOR_PLANE_REGS_H_
+#define _VS_CURSOR_PLANE_REGS_H_
+
+#include <linux/bits.h>
+
+#define VSDC_CURSOR_CONFIG(n) (0x1468 + 0x1080 * (n))
+#define VSDC_CURSOR_CONFIG_FMT_MASK GENMASK(1, 0)
+#define VSDC_CURSOR_CONFIG_FMT_ARGB8888 (0x2 << 0)
+#define VSDC_CURSOR_CONFIG_FMT_OFF (0x0 << 0)
+#define VSDC_CURSOR_CONFIG_IMG_UPDATE BIT(2)
+#define VSDC_CURSOR_CONFIG_COMMIT BIT(3)
+#define VSDC_CURSOR_CONFIG_SIZE_MASK GENMASK(7, 5)
+#define VSDC_CURSOR_CONFIG_SIZE_32 (0x0 << 5)
+#define VSDC_CURSOR_CONFIG_SIZE_64 (0x1 << 5)
+#define VSDC_CURSOR_CONFIG_SIZE_128 (0x2 << 5)
+#define VSDC_CURSOR_CONFIG_SIZE_256 (0x3 << 5)
+#define VSDC_CURSOR_CONFIG_Y_OFF_MASK GENMASK(12, 8)
+#define VSDC_CURSOR_CONFIG_Y_OFF(v) ((v) << 8)
+#define VSDC_CURSOR_CONFIG_X_OFF_MASK GENMASK(20, 16)
+#define VSDC_CURSOR_CONFIG_X_OFF(v) ((v) << 16)
+
+#define VSDC_CURSOR_ADDRESS(n) (0x146C + 0x1080 * (n))
+
+#define VSDC_CURSOR_LOCATION(n) (0x1470 + 0x1080 * (n))
+#define VSDC_CURSOR_LOCATION_X_MASK GENMASK(14, 0)
+#define VSDC_CURSOR_LOCATION_X(v) ((v) << 0)
+#define VSDC_CURSOR_LOCATION_Y_MASK GENMASK(30, 16)
+#define VSDC_CURSOR_LOCATION_Y(v) ((v) << 16)
+
+#define VSDC_CURSOR_BACKGROUND(n) (0x1474 + 0x1080 * (n))
+#define VSDC_CURSOR_BACKGRUOND_DEFAULT 0x00FFFFFF
+
+#define VSDC_CURSOR_FOREGROUND(n) (0x1478 + 0x1080 * (n))
+#define VSDC_CURSOR_FOREGRUOND_DEFAULT 0x00AAAAAA
+
+#endif /* _VS_CURSOR_PLANE_REGS_H_ */
diff --git a/drivers/gpu/drm/verisilicon/vs_plane.h b/drivers/gpu/drm/verisilicon/vs_plane.h
index 41875ea3d66a5..60b5b3a1bc22a 100644
--- a/drivers/gpu/drm/verisilicon/vs_plane.h
+++ b/drivers/gpu/drm/verisilicon/vs_plane.h
@@ -68,5 +68,6 @@ dma_addr_t vs_fb_get_dma_addr(struct drm_framebuffer *fb,
const struct drm_rect *src_rect);
struct drm_plane *vs_primary_plane_init(struct drm_device *dev, struct vs_dc *dc);
+struct drm_plane *vs_cursor_plane_init(struct drm_device *dev, struct vs_dc *dc);
#endif /* _VS_PLANE_H_ */
--
2.52.0
^ permalink raw reply related [flat|nested] 8+ messages in thread
* Claude review: drm: verisilicon: add hardware cursor support
2026-05-06 17:56 [PATCH v3 0/2] drm: verisilicon: add hardware cursor support Icenowy Zheng
2026-05-06 17:56 ` [PATCH v3 1/2] drm: verisilicon: add max cursor size to HWDB Icenowy Zheng
2026-05-06 17:56 ` [PATCH v3 2/2] drm: verisilicon: add support for cursor planes Icenowy Zheng
@ 2026-05-07 3:14 ` Claude Code Review Bot
2 siblings, 0 replies; 8+ messages in thread
From: Claude Code Review Bot @ 2026-05-07 3:14 UTC (permalink / raw)
To: dri-devel-reviews
Overall Series Review
Subject: drm: verisilicon: add hardware cursor support
Author: Icenowy Zheng <zhengxingda@iscas.ac.cn>
Patches: 3
Reviewed: 2026-05-07T13:14:48.302369
---
This is a well-structured 2-patch series adding hardware cursor support to the Verisilicon display controller driver. Patch 1 adds the HWDB plumbing, patch 2 implements the cursor plane. The approach is sound and the series shows good v3 iteration improvements (simpler validation, proper modifier list, visibility-based check skipping).
However, **patch 2 has a critical bug**: the negative-coordinate offset values are never shifted into their register field positions, so partially-offscreen cursors will not work correctly. There is also an off-by-one error in the coordinate range checks, minor typos in register defines, and a missing `cursor_width`/`cursor_height` setting in mode_config.
---
Generated by Claude Code Patch Reviewer
^ permalink raw reply [flat|nested] 8+ messages in thread
* Claude review: drm: verisilicon: add max cursor size to HWDB
2026-05-06 17:56 ` [PATCH v3 1/2] drm: verisilicon: add max cursor size to HWDB Icenowy Zheng
@ 2026-05-07 3:14 ` Claude Code Review Bot
0 siblings, 0 replies; 8+ messages in thread
From: Claude Code Review Bot @ 2026-05-07 3:14 UTC (permalink / raw)
To: dri-devel-reviews
Patch Review
This patch is clean and already has Thomas Zimmermann's Reviewed-by.
- The comment explaining the square-only constraint is good.
- Using `int32_t` for `max_cursor_size` matches how the cursor plane code uses it. Though `u32` would arguably be more natural for a size value, it's not wrong.
- All 4 HWDB entries are updated to `max_cursor_size = 64`. The cover letter says DC8200 variants support 32 and 64 while some DC8000 support up to 256, so 64 is a safe common value.
No issues with this patch.
---
Generated by Claude Code Patch Reviewer
^ permalink raw reply [flat|nested] 8+ messages in thread
* Claude review: drm: verisilicon: add support for cursor planes
2026-05-06 17:56 ` [PATCH v3 2/2] drm: verisilicon: add support for cursor planes Icenowy Zheng
@ 2026-05-07 3:14 ` Claude Code Review Bot
0 siblings, 0 replies; 8+ messages in thread
From: Claude Code Review Bot @ 2026-05-07 3:14 UTC (permalink / raw)
To: dri-devel-reviews
Patch Review
**Bug (critical) — Missing field shift macros for negative coordinate offsets:**
In `vs_cursor_plane_atomic_update()`, when writing negative X/Y offsets, the raw value is passed without shifting it to the register field position:
```c
regmap_update_bits(dc->regs, VSDC_CURSOR_CONFIG(output),
VSDC_CURSOR_CONFIG_X_OFF_MASK,
-state->crtc_x);
```
`VSDC_CURSOR_CONFIG_X_OFF_MASK` is `GENMASK(20, 16)` — it covers bits 16 through 20. The value `-state->crtc_x` is a small positive integer (1–31), which has no bits set in the 16–20 range. `regmap_update_bits` applies `val & mask`, so the offset will always be written as **zero**. The cursor will jump to position 0 instead of being partially offscreen.
The same bug applies to Y_OFF at `GENMASK(12, 8)`.
Fix: use the defined shift macros:
```c
VSDC_CURSOR_CONFIG_X_OFF(-state->crtc_x)
VSDC_CURSOR_CONFIG_Y_OFF(-state->crtc_y)
```
The header already defines `VSDC_CURSOR_CONFIG_X_OFF(v)` as `((v) << 16)` and `VSDC_CURSOR_CONFIG_Y_OFF(v)` as `((v) << 8)`, but they are never used.
**Bug (off-by-one) — Coordinate range checks allow values that overflow the register fields:**
```c
#define VSDC_CURSOR_LOCATION_MAX_POSITIVE BIT_MASK(15)
#define VSDC_CURSOR_LOCATION_MAX_NEGATIVE BIT_MASK(5)
```
`BIT_MASK(15)` = `1UL << 15` = 32768, but `VSDC_CURSOR_LOCATION_X_MASK` is `GENMASK(14, 0)` which is 15 bits holding values 0–32767. The check `coord <= 32768` accepts coord=32768, which cannot be represented in the field and would silently wrap to 0.
Similarly, `BIT_MASK(5)` = 32, but the 5-bit offset fields (`GENMASK(12, 8)` / `GENMASK(20, 16)`) hold values 0–31. The check `(-coord) <= 32` accepts 32, which overflows.
Both comparisons should use `<` instead of `<=`:
```c
return coord < VSDC_CURSOR_LOCATION_MAX_POSITIVE;
return (-coord) < VSDC_CURSOR_LOCATION_MAX_NEGATIVE;
```
**Issue — Missing `cursor_width`/`cursor_height` in mode_config:**
The driver never sets `drm->mode_config.cursor_width` or `drm->mode_config.cursor_height`. Userspace queries these via `DRM_CAP_CURSOR_WIDTH` / `DRM_CAP_CURSOR_HEIGHT`; they default to 64, which happens to match the current HWDB entries. But if a DC variant with `max_cursor_size = 256` is added later, userspace would still see 64. Consider setting these in `vs_mode_config_init()` or after the CRTCs are created.
**Nit — Typos in register header:**
```c
#define VSDC_CURSOR_BACKGRUOND_DEFAULT 0x00FFFFFF
#define VSDC_CURSOR_FOREGRUOND_DEFAULT 0x00AAAAAA
```
"BACKGRUOND" and "FOREGRUOND" should be "BACKGROUND" and "FOREGROUND". These macros are unused now but will rot if not fixed before someone references them.
**Nit — Garbled comment:**
```c
* values are unsigned. To for positive left-top coordinates the
```
"To for positive" doesn't parse. Probably should be "For positive". Also has a double space before "coordinates".
**Nit — License inconsistency:**
`vs_cursor_plane.c` uses `GPL-2.0-only` while `vs_cursor_plane_regs.h` uses `GPL-2.0-or-later`. Both are valid choices but the inconsistency within a single feature is worth noting.
**Minor — Unused `fb` variable when invisible:**
In `vs_cursor_plane_atomic_check()`, `fb` is assigned from `new_plane_state->fb` at the top but only used after the visibility check. This is harmless (no dereference on the invisible path) but could be moved down for clarity.
---
Generated by Claude Code Patch Reviewer
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2026-05-07 3:14 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-05-06 17:56 [PATCH v3 0/2] drm: verisilicon: add hardware cursor support Icenowy Zheng
2026-05-06 17:56 ` [PATCH v3 1/2] drm: verisilicon: add max cursor size to HWDB Icenowy Zheng
2026-05-07 3:14 ` Claude review: " Claude Code Review Bot
2026-05-06 17:56 ` [PATCH v3 2/2] drm: verisilicon: add support for cursor planes Icenowy Zheng
2026-05-07 3:14 ` Claude review: " Claude Code Review Bot
2026-05-07 3:14 ` Claude review: drm: verisilicon: add hardware cursor support Claude Code Review Bot
-- strict thread matches above, loose matches on Subject: below --
2026-03-24 6:02 [PATCH -next v2 0/2] " Icenowy Zheng
2026-03-24 21:03 ` Claude review: " Claude Code Review Bot
2026-03-09 8:53 [PATCH -next 0/2] " Icenowy Zheng
2026-03-10 2:36 ` Claude review: " Claude Code Review Bot
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox