在 2026-03-10二的 10:54 +0100,Thomas Zimmermann写道:
> Hi
> 
> Am 09.03.26 um 09:53 schrieb Icenowy Zheng:
> > Verisilicon display controllers support hardware cursors per output
> > port.
> > 
> > Add support for them as cursor planes.
> > 
> > Signed-off-by: Icenowy Zheng <[email protected]>
> > ---
> >   drivers/gpu/drm/verisilicon/Makefile          |   3 +-
> >   drivers/gpu/drm/verisilicon/vs_crtc.c         |  11 +-
> >   drivers/gpu/drm/verisilicon/vs_cursor_plane.c | 273
> > ++++++++++++++++++
> >   .../drm/verisilicon/vs_cursor_plane_regs.h    |  44 +++
> >   drivers/gpu/drm/verisilicon/vs_plane.h        |   1 +
> >   .../drm/verisilicon/vs_primary_plane_regs.h   |   5 +-
> >   6 files changed, 333 insertions(+), 4 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..4c86519458077
> > --- /dev/null
> > +++ b/drivers/gpu/drm/verisilicon/vs_cursor_plane.c
> > @@ -0,0 +1,273 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Copyright (C) 2026 Institute of Software, Chinese Academy of
> > Sciences (ISCAS)
> > + *
> > + * Authors:
> > + * Icenowy Zheng <[email protected]>
> > + */
> > +
> > +#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_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 (!crtc)
> 
> Use "if (!new_plane_state->visible)"  for this test. The plane might
> be 
> invisible if the crtc has been set.  you should return in all such
> cases 
> unless you have a good reason not to.
> 
> > +           return 0; /* Skip validity check */
> > +
> > +   vcrtc = drm_crtc_to_vs_crtc(crtc);
> > +   dc = vcrtc->dc;
> > +
> > +   /* Only certain square sizes is supported */
> > +   switch (new_plane_state->crtc_w) {
> > +   case 32:
> > +   case 64:
> > +   case 128:
> > +   case 256:
> 
> Instead of using max_cursor_size, could this be a HW-specific bit
> field 
> that you test the cursor size against? Like
> 
> if (!is_power_of_2(crtc_w) || !(max_cursor_size & crtc_w)
>      return -EINVAL
> 
> But also see my comment on the switch in atomic_update.
> 
> 
> > +           break;
> > +   default:
> > +           return -EINVAL;
> > +   }
> > +
> > +   if (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;
> > +
> > +   if (fb) {
> > +           /* Only ARGB8888 is supported */
> > +           if (drm_WARN_ON_ONCE(plane->dev,
> > +                                fb->format->format !=
> > DRM_FORMAT_ARGB8888))
> > +                   return -EINVAL;
> 
> We already do this check at [1], including supported modifiers.
> 
> [1] 
> https://elixir.bootlin.com/linux/v6.19.6/source/drivers/gpu/drm/drm_atomic.c#L739
> 
> > +
> > +           /* 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;
> > +   }
> 
> You already set atomic_disable in the plane_helper_funcs. Therefore
> you 
> will never take this conditional.

Well it looks like the condition to call atomic_disable instead of
atomic_update is drm_atomic_plane_disabling(), which only checks
whether the plane has a CRTC bound before and no CRTC bound after.

Here the code needs to be here to handle the situation that the plane
is bound but not visible, like what's mentioned in your previous
comment of atomic_check .

Thanks,
Icenowy

> 
> > +
> > +   vcrtc = drm_crtc_to_vs_crtc(crtc);
> > +   output = vcrtc->id;
> > +   dc = vcrtc->dc;
> > +
> > +   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;
> > +   }
> 
> Here is another case where you can move a potential error state into
> the 
> atomic_check.  If you add a custom vc_cursor_plane_state, you can add
> a 
> size field that stores the size constants.  Then it would make sense
> to 
> keep the switch statement in atomic check and set the field right 
> there.  Here in atomic_update, you'd need no switch.
> 
> > +
> > +   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,
> > +};
> 
> Maybe also add
> 
> static const u64 vs_cursor_plane_format_modifiers[] = {
>    DRM_FORMAT_MOD_LINEAR,
>    DRM_FORMAT_MOD_INVALID,
> };
> 
> and pass it to drmm_universal_plane_alloc().  Using NULL there is 
> supported, but I'd consider it bad style. Your choice.
> 
> Best regards
> Thomas
> 
> > +
> > +struct drm_plane *vs_cursor_plane_init(struct drm_device *drm_dev,
> > +                                  struct vs_dc *dc)
> > +{
> > +   struct drm_plane *plane;
> > +
> > +   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),
> > +                                      NULL,
> > +                                      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 <[email protected]>
> > + *
> > + * 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_ */
> > diff --git a/drivers/gpu/drm/verisilicon/vs_primary_plane_regs.h
> > b/drivers/gpu/drm/verisilicon/vs_primary_plane_regs.h
> > index cbb125c46b390..61a48c2faa1b2 100644
> > --- a/drivers/gpu/drm/verisilicon/vs_primary_plane_regs.h
> > +++ b/drivers/gpu/drm/verisilicon/vs_primary_plane_regs.h
> > @@ -1,9 +1,12 @@
> >   /* SPDX-License-Identifier: GPL-2.0-only */
> >   /*
> > - * Copyright (C) 2025 Icenowy Zheng <[email protected]>
> > + * Copyright (C) 2026 Institute of Software, Chinese Academy of
> > Sciences (ISCAS)
> >    *
> >    * Based on vs_dc_hw.h, which is:
> >    *   Copyright (C) 2023 VeriSilicon Holdings Co., Ltd.
> > + *
> > + * Authors:
> > + * Icenowy Zheng <[email protected]>
> >    */
> >   
> >   #ifndef _VS_PRIMARY_PLANE_REGS_H_

Reply via email to