在 2025-08-16星期六的 10:04 +0000，Yao Zi写道：
> On Fri, Aug 15, 2025 at 12:40:43AM +0800, Icenowy Zheng wrote:
> > This is a from-scratch driver targeting Verisilicon DC-series
> > display
> > controllers, which feature self-identification functionality like
> > their
> > GC-series GPUs.
> >
> > Only DC8200 is being supported now, and only the main framebuffer
> > is set
> > up (as the DRM primary plane). Support for more DC models and more
> > features is my further targets.
> >
> > As the display controller is delivered to SoC vendors as a whole
> > part,
> > this driver does not use component framework and extra bridges
> > inside a
> > SoC is expected to be implemented as dedicated bridges (this driver
> > properly supports bridge chaining).
> >
> > Signed-off-by: Icenowy Zheng <u...@icenowy.me>
>
> Thank you for the great work!
>
> > ---
> > drivers/gpu/drm/Kconfig | 2 +
> > drivers/gpu/drm/Makefile | 1 +
> > drivers/gpu/drm/verisilicon/Kconfig | 15 +
> > drivers/gpu/drm/verisilicon/Makefile | 5 +
> > drivers/gpu/drm/verisilicon/vs_bridge.c | 330
> > ++++++++++++++++++
> > drivers/gpu/drm/verisilicon/vs_bridge.h | 40 +++
> > drivers/gpu/drm/verisilicon/vs_bridge_regs.h | 47 +++
> > drivers/gpu/drm/verisilicon/vs_crtc.c | 217 ++++++++++++
> > drivers/gpu/drm/verisilicon/vs_crtc.h | 29 ++
> > drivers/gpu/drm/verisilicon/vs_crtc_regs.h | 60 ++++
> > drivers/gpu/drm/verisilicon/vs_dc.c | 233 +++++++++++++
> > drivers/gpu/drm/verisilicon/vs_dc.h | 39 +++
> > drivers/gpu/drm/verisilicon/vs_dc_top_regs.h | 27 ++
> > drivers/gpu/drm/verisilicon/vs_drm.c | 177 ++++++++++
> > drivers/gpu/drm/verisilicon/vs_drm.h | 29 ++
> > drivers/gpu/drm/verisilicon/vs_hwdb.c | 150 ++++++++
> > drivers/gpu/drm/verisilicon/vs_hwdb.h | 29 ++
> > drivers/gpu/drm/verisilicon/vs_plane.c | 102 ++++++
> > drivers/gpu/drm/verisilicon/vs_plane.h | 68 ++++
> > .../gpu/drm/verisilicon/vs_primary_plane.c | 166 +++++++++
> > .../drm/verisilicon/vs_primary_plane_regs.h | 53 +++
> > 21 files changed, 1819 insertions(+)
> > create mode 100644 drivers/gpu/drm/verisilicon/Kconfig
> > create mode 100644 drivers/gpu/drm/verisilicon/Makefile
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_bridge.c
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_bridge.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_bridge_regs.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_crtc.c
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_crtc.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_crtc_regs.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_dc.c
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_dc.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_dc_top_regs.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_drm.c
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_drm.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_hwdb.c
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_hwdb.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_plane.c
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_plane.h
> > create mode 100644 drivers/gpu/drm/verisilicon/vs_primary_plane.c
> > create mode 100644
> > drivers/gpu/drm/verisilicon/vs_primary_plane_regs.h
> >
>
> ...
>
> > diff --git a/drivers/gpu/drm/verisilicon/vs_bridge.c
> > b/drivers/gpu/drm/verisilicon/vs_bridge.c
> > new file mode 100644
> > index 0000000000000..c8caf31fac7d6
> > --- /dev/null
> > +++ b/drivers/gpu/drm/verisilicon/vs_bridge.c
>
> ...
>
> > +static bool vs_bridge_out_dp_fmt_supported(u32 out_fmt)
> > +{
> > + unsigned int i;
> > +
> > + for (i = 0; i < ARRAY_SIZE(vsdc_dp_supported_fmts); i++)
> > + if (vsdc_dp_supported_fmts[i].linux_fmt == out_fmt)
> > + break;
> > +
> > + return !(i == ARRAY_SIZE(vsdc_dp_supported_fmts));
>
> You could simplify the return statement by returning early in the
> loop
> if out_fmt matches.
Thanks for this tip, this was originally coded in two different
functions, but then extracted to be such a common function.
>
> > +}
>
> > +static int vs_bridge_detect_output_interface(struct device_node
> > *of_node,
> > + unsigned int output)
> > +{
> > + int ret;
> > + struct device_node *remote;
> > +
> > + remote = of_graph_get_remote_node(of_node, output,
> > +
> > VSDC_OUTPUT_INTERFACE_DPI);
> > + if (remote) {
> > + ret = VSDC_OUTPUT_INTERFACE_DPI;
> > + } else {
> > + remote = of_graph_get_remote_node(of_node, output,
> > +
> > VSDC_OUTPUT_INTERFACE_DP);
> > + if (remote)
> > + ret = VSDC_OUTPUT_INTERFACE_DP;
> > + else
> > + ret = -ENODEV;
> > + }
>
> remote is leaked here, it should be released with of_node_put IMHO.
Thanks for this, will fix it.
>
> > + return ret;
> > +}
> > +
> > +struct vs_bridge *vs_bridge_init(struct drm_device *drm_dev,
> > + struct vs_crtc *crtc)
> > +{
> > + unsigned int output = crtc->id;
> > + struct vs_bridge *bridge;
> > + struct drm_bridge *next;
> > + enum vs_bridge_output_interface intf;
> > + int ret;
>
> ...
>
> > + bridge = devm_kzalloc(drm_dev->dev, sizeof(*bridge),
> > GFP_KERNEL);
> > + if (!bridge)
> > + return ERR_PTR(-ENOMEM);
>
> ...
>
> > +err_cleanup_encoder:
> > + drm_encoder_cleanup(&bridge->enc);
> > +err_free_bridge:
> > + devm_kfree(drm_dev->dev, bridge);
>
> Though is technically correct, this devm_kfree() is unnecessary.
> Resources registered with devres are automatically released when
> probing
> fails[1][2].
>
> > + return ERR_PTR(ret);
> > +}
>
> ...
>
> > diff --git a/drivers/gpu/drm/verisilicon/vs_crtc.c
> > b/drivers/gpu/drm/verisilicon/vs_crtc.c
> > new file mode 100644
> > index 0000000000000..46c4191b82f49
> > --- /dev/null
> > +++ b/drivers/gpu/drm/verisilicon/vs_crtc.c
> > @@ -0,0 +1,217 @@
>
> ...
>
> > +static enum drm_mode_status
> > +vs_crtc_mode_valid(struct drm_crtc *crtc, const struct
> > drm_display_mode *mode)
> > +{
> > + struct vs_crtc *vcrtc = drm_crtc_to_vs_crtc(crtc);
> > + struct vs_dc *dc = vcrtc->dc;
> > + unsigned int output = vcrtc->id;
> > + long rate;
> > +
> > + if (mode->htotal > 0x7FFF)
> > + return MODE_BAD_HVALUE;
> > + if (mode->vtotal > 0x7FFF)
> > + return MODE_BAD_VVALUE;
> > +
> > + rate = clk_round_rate(dc->pix_clk[output], mode->clock *
> > 1000);
> > + if (rate <= 0)
> > + return MODE_CLOCK_RANGE;
>
> Some round_rate implementations may not return zero or error on an
> unsupported clock rate, instead they simply return a value differing
> a
> lot from the requested rate. Thus I think you should also check
> whether
> the resulted rate is acceptable as well, which applies for
> vs_crtc_mode_fixup() as well.
Well I don't know how to decide a round rate is good (sometimes the
clock might be a little off but setting it is still better than
rejecting the output device), so I think this should be considered a
further problem.
>
> > + return MODE_OK;
> > +}
> > +
> > +static bool vs_crtc_mode_fixup(struct drm_crtc *crtc,
> > + const struct drm_display_mode *m,
> > + struct drm_display_mode
> > *adjusted_mode)
> > +{
> > + struct vs_crtc *vcrtc = drm_crtc_to_vs_crtc(crtc);
> > + struct vs_dc *dc = vcrtc->dc;
> > + unsigned int output = vcrtc->id;
> > + long clk_rate;
> > +
> > + drm_mode_set_crtcinfo(adjusted_mode, 0);
> > +
> > + /* Feedback the pixel clock to crtc_clock */
> > + clk_rate = adjusted_mode->crtc_clock * 1000;
> > + clk_rate = clk_round_rate(dc->pix_clk[output], clk_rate);
> > + if (clk_rate <= 0)
> > + return false;
> > +
> > + adjusted_mode->crtc_clock = clk_rate / 1000;
> > +
> > + return true;
> > +}
>
> ...
>
> > +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;
> > + int ret;
> > +
> > + vcrtc = devm_kzalloc(drm_dev->dev, sizeof(*vcrtc),
> > GFP_KERNEL);
> > + if (!vcrtc)
> > + return ERR_PTR(-ENOMEM);
> > + vcrtc->dc = dc;
> > + vcrtc->id = output;
> > +
> > + /* Create our primary plane */
> > + primary = vs_primary_plane_init(drm_dev, dc);
> > + if (IS_ERR(primary)) {
> > + dev_err(drm_dev->dev, "Couldn't create the primary
> > plane\n");
> > + return ERR_PTR(PTR_ERR(primary));
> > + }
>
> Suggest dev_err_probe which handles deferred probing as well.
Well I don't think CRTC/Plane initialization will lead to deferred
probing.
>
> > +
> > + ret = drm_crtc_init_with_planes(drm_dev, &vcrtc->base,
> > + primary,
> > + NULL,
> > + &vs_crtc_funcs,
> > + NULL);
> > + if (ret) {
> > + dev_err(drm_dev->dev, "Couldn't initialize
> > CRTC\n");
> > + return ERR_PTR(ret);
> > + }
> > +
> > + drm_crtc_helper_add(&vcrtc->base, &vs_crtc_helper_funcs);
> > +
> > + return vcrtc;
> > +}
>
> ...
>
> > diff --git a/drivers/gpu/drm/verisilicon/vs_dc.c
> > b/drivers/gpu/drm/verisilicon/vs_dc.c
> > new file mode 100644
> > index 0000000000000..98384559568c4
> > --- /dev/null
> > +++ b/drivers/gpu/drm/verisilicon/vs_dc.c
> > @@ -0,0 +1,233 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Copyright (C) 2025 Icenowy Zheng <u...@icenowy.me>
> > + */
> > +
> > +#include <linux/dma-mapping.h>
> > +#include <linux/of.h>
> > +#include <linux/of_graph.h>
> > +
> > +#include "vs_crtc.h"
> > +#include "vs_dc.h"
> > +#include "vs_dc_top_regs.h"
> > +#include "vs_drm.h"
> > +#include "vs_hwdb.h"
> > +
> > +static const struct regmap_config vs_dc_regmap_cfg = {
> > + .reg_bits = 32,
> > + .val_bits = 32,
> > + .reg_stride = sizeof(u32),
> > + /* VSDC_OVL_CONFIG_EX(1) */
> > + .max_register = 0x2544,
> > + .cache_type = REGCACHE_NONE,
>
> I think cache_type = REGCACHE_NONE is the default, thus it should be
> okay to omitted it?
Sounds right. Thanks.
>
> > +};
>
> ...
>
> > +static int vs_dc_probe(struct platform_device *pdev)
> > +{
>
> ...
>
> > + dc->core_clk = devm_clk_get(dev, "core");
> > + if (IS_ERR(dc->core_clk)) {
> > + dev_err(dev, "can't get core clock\n");
> > + return PTR_ERR(dc->core_clk);
> > + }
> > +
> > + dc->axi_clk = devm_clk_get(dev, "axi");
> > + if (IS_ERR(dc->axi_clk)) {
> > + dev_err(dev, "can't get axi clock\n");
> > + return PTR_ERR(dc->axi_clk);
> > + }
> > +
> > + dc->ahb_clk = devm_clk_get(dev, "ahb");
> > + if (IS_ERR(dc->ahb_clk)) {
> > + dev_err(dev, "can't get ahb clock\n");
> > + return PTR_ERR(dc->ahb_clk);
> > + }
>
> I think devm_clk_get_enabled() will help here.
Thanks for the tip, I don't know why I forgot it here (because I used
devm_clk_get_enabled in th1520-dw-hdmi driver).
>
> > + for (i = 0; i < outputs; i++) {
> > + snprintf(pixclk_name, sizeof(pixclk_name), "pix%u",
> > i);
> > + dc->pix_clk[i] = devm_clk_get(dev, pixclk_name);
> > + if (IS_ERR(dc->pix_clk[i])) {
> > + dev_err(dev, "can't get pixel clk %u\n",
> > i);
> > + return PTR_ERR(dc->pix_clk[i]);
> > + }
> > + }
>
> ...
>
> > +err_ahb_clk_disable:
> > + clk_disable_unprepare(dc->ahb_clk);
> > +err_axi_clk_disable:
> > + clk_disable_unprepare(dc->axi_clk);
> > +err_core_clk_disable:
> > + clk_disable_unprepare(dc->core_clk);
>
> And you could get avoid these clk_disable_unprepare(). Also the ones
> in
> the remove callback.
>
> > +err_rst_assert:
> > + reset_control_bulk_assert(VSDC_RESET_COUNT, dc->rsts);
> > + return ret;
> > +}
> > +
> > +static void vs_dc_remove(struct platform_device *pdev)
> > +{
> > + struct vs_dc *dc = dev_get_drvdata(&pdev->dev);
> > +
> > + vs_drm_finalize(dc);
> > +
> > + dev_set_drvdata(&pdev->dev, NULL);
> > +
> > + clk_disable_unprepare(dc->ahb_clk);
> > + clk_disable_unprepare(dc->axi_clk);
> > + clk_disable_unprepare(dc->core_clk);
> > + reset_control_bulk_assert(VSDC_RESET_COUNT, dc->rsts);
> > +}
>
> ...
>
> > diff --git a/drivers/gpu/drm/verisilicon/vs_primary_plane.c
> > b/drivers/gpu/drm/verisilicon/vs_primary_plane.c
> > new file mode 100644
> > index 0000000000000..25d6e01cc8b71
> > --- /dev/null
> > +++ b/drivers/gpu/drm/verisilicon/vs_primary_plane.c
> > @@ -0,0 +1,166 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Copyright (C) 2025 Icenowy Zheng <u...@icenowy.me>
> > + */
> > +
> > +#include <linux/regmap.h>
> > +
> > +#include <drm/drm_atomic.h>
> > +#include <drm/drm_atomic_helper.h>
> > +#include <drm/drm_crtc.h>
> > +#include <drm/drm_fb_dma_helper.h>
> > +#include <drm/drm_fourcc.h>
> > +#include <drm/drm_framebuffer.h>
> > +#include <drm/drm_gem_atomic_helper.h>
> > +#include <drm/drm_gem_dma_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_primary_plane_regs.h"
> > +
> > +static int vs_primary_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_crtc_state *crtc_state;
> > +
> > + if (!crtc)
> > + return 0;
> > +
> > + crtc_state = drm_atomic_get_existing_crtc_state(state,
> > + crtc);
>
> I think these arguments could be merged into a single line.
>
> > + if (WARN_ON(!crtc_state))
> > + return -EINVAL;
> > +
> > + return drm_atomic_helper_check_plane_state(new_plane_state,
> > + crtc_state,
> > +
> > DRM_PLANE_NO_SCALING,
> > +
> > DRM_PLANE_NO_SCALING,
> > + false, true);
> > +}
>
> ...
>
> > +struct drm_plane *vs_primary_plane_init(struct drm_device
> > *drm_dev, struct vs_dc *dc)
> > +{
> > + struct drm_plane *plane;
> > + int ret;
> > +
> > + plane = devm_kzalloc(drm_dev->dev, sizeof(*plane),
> > GFP_KERNEL);
> > + if (!plane)
> > + return ERR_PTR(-ENOMEM);
> > +
> > + ret = drm_universal_plane_init(drm_dev, plane, 0,
> > + &vs_primary_plane_funcs,
> > + dc->identity.formats->array,
> > + dc->identity.formats->num,
> > + NULL,
> > + DRM_PLANE_TYPE_PRIMARY,
> > + NULL);
> > +
> > + if (ret) {
> > + devm_kfree(drm_dev->dev, plane);
>
> Similar to vs_bridge_init's case, this devm_kfree() isn't
> unnecessary,
> either.
>
> > + return ERR_PTR(ret);
> > + }
> > +
> > + drm_plane_helper_add(plane,
> > &vs_primary_plane_helper_funcs);
> > +
> > + return plane;
> > +}
>
> Best regards,
> Yao Zi
>
> [1]:
> https://elixir.bootlin.com/linux/v6.16/source/drivers/base/dd.c#L723
> [2]:
> https://elixir.bootlin.com/linux/v6.16/source/drivers/base/dd.c#L549
