Re: [PATCH v2 14/19] drm: rcar-du: Add support for CMM
From: Laurent Pinchart
Date: Thu Aug 22 2019 - 11:11:10 EST
Hi Jacopo,
On Thu, Aug 22, 2019 at 03:01:46PM +0200, Jacopo Mondi wrote:
> On Tue, Aug 20, 2019 at 08:34:44PM +0300, Laurent Pinchart wrote:
> > On Sat, Jul 06, 2019 at 04:07:41PM +0200, Jacopo Mondi wrote:
> >> Add a driver for the R-Car Display Unit Color Correction Module.
> >>
> >> In most of Gen3 SoCs, each DU output channel is provided with a CMM unit
> >> to perform image enhancement and color correction.
> >>
> >> Add support for CMM through a driver that supports configuration of
> >> the 1-dimensional LUT table. More advanced CMM feature will be
> >> implemented on top of this basic one.
> >>
> >> Signed-off-by: Jacopo Mondi <jacopo+renesas@xxxxxxxxxx>
> >> ---
> >> drivers/gpu/drm/rcar-du/Kconfig | 7 +
> >> drivers/gpu/drm/rcar-du/Makefile | 1 +
> >> drivers/gpu/drm/rcar-du/rcar_cmm.c | 292 +++++++++++++++++++++++++++++
> >> drivers/gpu/drm/rcar-du/rcar_cmm.h | 38 ++++
> >> 4 files changed, 338 insertions(+)
> >> create mode 100644 drivers/gpu/drm/rcar-du/rcar_cmm.c
> >> create mode 100644 drivers/gpu/drm/rcar-du/rcar_cmm.h
> >>
> >> diff --git a/drivers/gpu/drm/rcar-du/Kconfig b/drivers/gpu/drm/rcar-du/Kconfig
> >> index 1529849e217e..539d232790d1 100644
> >> --- a/drivers/gpu/drm/rcar-du/Kconfig
> >> +++ b/drivers/gpu/drm/rcar-du/Kconfig
> >> @@ -13,6 +13,13 @@ config DRM_RCAR_DU
> >> Choose this option if you have an R-Car chipset.
> >> If M is selected the module will be called rcar-du-drm.
> >>
> >> +config DRM_RCAR_CMM
> >> + bool "R-Car DU Color Management Module (CMM) Support"
> >> + depends on DRM && OF
> >> + depends on DRM_RCAR_DU
> >> + help
> >> + Enable support for R-Car Color Management Module (CMM).
> >> +
> >> config DRM_RCAR_DW_HDMI
> >> tristate "R-Car DU Gen3 HDMI Encoder Support"
> >> depends on DRM && OF
> >> diff --git a/drivers/gpu/drm/rcar-du/Makefile b/drivers/gpu/drm/rcar-du/Makefile
> >> index 6c2ed9c46467..4d1187ccc3e5 100644
> >> --- a/drivers/gpu/drm/rcar-du/Makefile
> >> +++ b/drivers/gpu/drm/rcar-du/Makefile
> >> @@ -15,6 +15,7 @@ rcar-du-drm-$(CONFIG_DRM_RCAR_LVDS) += rcar_du_of.o \
> >> rcar-du-drm-$(CONFIG_DRM_RCAR_VSP) += rcar_du_vsp.o
> >> rcar-du-drm-$(CONFIG_DRM_RCAR_WRITEBACK) += rcar_du_writeback.o
> >>
> >> +obj-$(CONFIG_DRM_RCAR_CMM) += rcar_cmm.o
> >> obj-$(CONFIG_DRM_RCAR_DU) += rcar-du-drm.o
> >> obj-$(CONFIG_DRM_RCAR_DW_HDMI) += rcar_dw_hdmi.o
> >> obj-$(CONFIG_DRM_RCAR_LVDS) += rcar_lvds.o
> >> diff --git a/drivers/gpu/drm/rcar-du/rcar_cmm.c b/drivers/gpu/drm/rcar-du/rcar_cmm.c
> >> new file mode 100644
> >> index 000000000000..76ed3fce2b33
> >> --- /dev/null
> >> +++ b/drivers/gpu/drm/rcar-du/rcar_cmm.c
> >> @@ -0,0 +1,292 @@
> >> +// SPDX-License-Identifier: GPL-2.0+
> >> +/*
> >> + * rcar_cmm.c -- R-Car Display Unit Color Management Module
> >> + *
> >> + * Copyright (C) 2019 Jacopo Mondi <jacopo+renesas@xxxxxxxxxx>
> >> + */
> >> +
> >> +#include <linux/clk.h>
> >> +#include <linux/io.h>
> >> +#include <linux/module.h>
> >> +#include <linux/of.h>
> >> +#include <linux/platform_device.h>
> >> +#include <linux/pm.h>
> >> +
> >> +#include <drm/drm_atomic.h>
> >
> > The only thing you need from DRM is the drm_color_lut structure, so I
> > would include drm/drm_mode.h instead.
> >
> >> +#include "rcar_cmm.h"
> >> +
> >> +#define CM2_LUT_CTRL 0x0000
> >> +#define CM2_LUT_CTRL_EN BIT(0)
> >
> > The datasheet names the bit LUT_EN, I would thus rename the macro to
> > CM2_LUT_CTRL_LUT_EN.
> >
> >> +#define CM2_LUT_TBLA_BASE 0x0600
> >> +#define CM2_LUT_TBLA(__i) (CM2_LUT_TBLA_BASE + (__i) * 4)
> >
> > Similarly, the datasheet names the register CM2_LUT_TBL (and the LUT
> > table B is named CM2_LUT_TBL2), would it make sense to stick to those
> > names ?
>
> Ack on all of these
>
> >> +
> >> +struct rcar_cmm {
> >> + struct clk *clk;
> >> + void __iomem *base;
> >> + bool enabled;
> >> +
> >> + /*
> >> + * restore: LUT restore flag
> >
> > I'm none the wiser after reading this comment :-)
> >
> >> + * running: LUT operating flag
> >> + * size: Number of programmed entries in the LUT table
> >> + * table: Scratch buffer where to store the LUT table entries to be
> >> + * later restored.
> >
> > If you want to document individual fields I propose using kerneldoc
> > syntax.
> >
> > * @lut.restore: ...
> > ...
>
> Yeah, might be a bit of over-documentation here. I don't mind it to be
> honest, but I'm fine if someone wants this to be dropped.
I think it's important to document all fields.
> >> + */
> >> + struct {
> >> + bool restore;
> >> + bool running;
> >> + unsigned int size;
> >> + struct drm_color_lut table[CMM_GAMMA_LUT_SIZE];
> >> + } lut;
> >
> > I think the lut.running field name is a bit confusing, as you modify it
> > based on the lut.enable field in the cmm config structure. I would name
> > it lut.enabled. That could then possibly be confused with cmm.enabled
> > (although in my opinion that's fine), if you're concerned by that I
> > would rename that field to running. It would be more logical to consider
> > the CMM as a whole as running or stopped, with the LUT (and later the
> > CLU) enabled or disabled.
>
> I'm a bit bothered that we would have
> rcmm.enabled
> rcmm.lut.enabled
> rcmm_config.lut.enable
>
> I'll see how it looks. enabled is probably the right name for all of
> these fields, but it might get confusing...
As long as we keep cmm->enabled and cmm->lut.enabled in the code (and
don't create alias local variables for cmm of cmm->lut with confusing
names such as dev for instance) I think it will be OK.
> >> +};
> >> +
> >> +static inline int rcar_cmm_read(struct rcar_cmm *rcmm, u32 reg)
> >> +{
> >> + return ioread32(rcmm->base + reg);
> >> +}
> >> +
> >> +static inline void rcar_cmm_write(struct rcar_cmm *rcmm, u32 reg, u32 data)
> >> +{
> >> + iowrite32(data, rcmm->base + reg);
> >> +}
> >> +
> >> +static void rcar_cmm_lut_write(struct rcar_cmm *rcmm, unsigned int size,
> >
> > s/unsigned int/size_t/ ?
> >
> >> + struct drm_color_lut *lut)
> >
> > You can make this pointer const.
>
> Ack
>
> >> +{
> >> + unsigned int i;
> >> +
> >> + for (i = 0; i < size; ++i) {
> >> + struct drm_color_lut *entry = &lut[i];
> >> +
> >> + u32 val = (entry->red & 0xff) << 16 |
> >> + (entry->green & 0xff) << 8 |
> >> + (entry->blue & 0xff);
> >> + rcar_cmm_write(rcmm, CM2_LUT_TBLA(i), val);
> >> + }
> >> +}
> >> +
> >> +/**
> >> + * rcar_cmm_setup() - configure the CMM unit
> >> + *
> >> + * @pdev: The platform device associated with the CMM instance
> >> + * @config: The CRTC provided configuration.
> >> + *
> >> + * Configure the CMM unit with the CRTC provided configuration.
> >
> > s/CRTC provided/CRTC-provided/
> >
> > "CRTC-provided" is a compound adjective, qualifying the noun
> > "configuration". It thus needs a hyphen. If you had written "The process
> > uses the CRTC provided to this function", then no hyphen would be
> > needed, as "provided" then qualifies the noun "CRTC", without the
> > combination being used as an adjective.
> >
> >> + * Currently enabling, disabling and programming of the 1-D LUT unit is
> >> + * supported.
> >> + */
> >> +int rcar_cmm_setup(struct platform_device *pdev,
> >> + const struct rcar_cmm_config *config)
> >> +{
> >> + struct rcar_cmm *rcmm = platform_get_drvdata(pdev);
> >> + unsigned int i;
> >> +
> >> + if (config->lut.size > CMM_GAMMA_LUT_SIZE)
> >> + return -EINVAL;
> >> +
> >> + /*
> >> + * As cmm_setup is called by atomic commit tail helper, it might be
> >
> > s/cmm_setup/rcar_cmm_setup()/ (or just "this function").
> >
> >> + * called before enabling the CRTC (which calls cmm_enable()).
> >
> > I would phrase this as "... it might be called when the CMM is disabled.
> > We can't program the hardware in that case, store the configuration
> > internally and apply it when the CMM will be enabled by the CRTC through
> > by rcar_cmm_enable()." and remove the next comment.
>
> Ack
>
> >> + */
> >> + if (!rcmm->enabled) {
> >> + if (!config->lut.enable)
> >> + return 0;
> >> +
> >> + /*
> >> + * Store the LUT table entries in the scratch buffer to be later
> >> + * programmed at enable time.
> >> + */
> >> + for (i = 0; i < config->lut.size; ++i)
> >> + rcmm->lut.table[i] = config->lut.table[i];
> >
> > Can you do a memcpy() over the whole table ?
> >
> > memcpy(rcmm->lut.table, config->lut.table,
> > sizeof(*rcmm->lut.table) * config.lut.size);
>
> Yeah, probably better
>
> >> +
> >> + rcmm->lut.size = config->lut.size;
> >> + rcmm->lut.restore = true;
> >> +
> >> + return 0;
> >> + }
> >> +
> >> + /* Stop LUT operations, if requested. */
> >> + if (rcmm->lut.running && !config->lut.enable) {
> >> + rcar_cmm_write(rcmm, CM2_LUT_CTRL, 0);
> >> + rcmm->lut.running = 0;
> >> + rcmm->lut.size = 0;
> >> +
> >> + return 0;
> >> + }
> >> +
> >> + /* Enable LUT and program the new gamma table values. */
> >> + if (!rcmm->lut.running) {
> >
> > Should this be !rcmm->lut.running && config->lut.enable ? Or do you rely
> > on the caller to not try to disable the LUT when it's not currently
> > enabled ? If you rely on the caller than I think you should rely on it
> > for the enabling case above too, and write is if (!config->lut.enabled).
> > Otherwise I think you're mishandling the !running && !enable, it will
> > end up enabling the LUT.
>
> I think it's fine, as if (!rcmm->enable) then (!rcmm->lut.running) so
> the (!running && !enable) you have pointed out gets caught by the very
> first condition check here above (!rcmm->enabled).
>
> I'll try to restructure this to be more readable and as you suggested
> get rid of the restore field.
>
> >> + rcar_cmm_write(rcmm, CM2_LUT_CTRL, CM2_LUT_CTRL_EN);
> >> + rcmm->lut.running = true;
> >> + }
> >> +
> >> + rcar_cmm_lut_write(rcmm, config->lut.size, config->lut.table);
> >
> > I'm still very puzzled by the fact that you have to write the LUT
> > contents after enabling the LUT. The datasheet states
>
> I know -\_(;.;)_/- (is this the crying cthulhu emoticon ?)
>
> > "Note that if the module that references that space is operating, read
> > and write accesses to the relevant space are prohibited. In case of
> > double buffer mode, referenced side of LUT is distinguished by
> > CM2_CTL1.BFS."
> >
> > It looks to me like you may have to implement double-buffering, but even
> > without that,
>
> I think you have left out the end of the sentence, but I agree that
> what the manual reports suggests the table should be programmed when
> it is not operating, but it also mentions double buffering, so in case
> we're using single buffer mode maybe this does not apply?
I think it does, I'm sorry :-)
> with double buffering this is going to change anyway, but so far,
> that's the only reliable operation sequence I have found...
Then please add a FIXME comment explaining that this is weird and needs
to be figured out.
> >> + rcmm->lut.size = config->lut.size;
> >> +
> >> + return 0;
> >> +}
> >> +EXPORT_SYMBOL_GPL(rcar_cmm_setup);
> >> +
> >> +/**
> >> + * rcar_cmm_enable - enable the CMM unit
> >> + *
> >> + * @pdev: The platform device associated with the CMM instance
> >> + *
> >> + * Enable the CMM unit by enabling the parent clock and enabling the CMM
> >> + * components, such as 1-D LUT, if requested.
> >> + */
> >> +int rcar_cmm_enable(struct platform_device *pdev)
> >> +{
> >> + struct rcar_cmm *rcmm = platform_get_drvdata(pdev);
> >> + int ret;
> >> +
> >> + if (!rcmm)
> >> + return -EPROBE_DEFER;
> >> +
> >> + ret = clk_prepare_enable(rcmm->clk);
> >> + if (ret)
> >> + return ret;
> >> +
> >
> > Didn't you say this version would use runtime PM ? :-)
>
> Ups. I do for suspend/resume not for runtime_suspend/resume. It will
> be trivial to add.
>
> >> + /* Apply the LUT table values saved at cmm_setup time. */
> >
> > rcar_cmm_setup() here too.
> >
> >> + if (rcmm->lut.restore) {
> >> + rcar_cmm_write(rcmm, CM2_LUT_CTRL, CM2_LUT_CTRL_EN);
> >> + rcar_cmm_lut_write(rcmm, rcmm->lut.size, rcmm->lut.table);
> >> +
> >> + rcmm->lut.restore = false;
> >> + rcmm->lut.running = true;
> >> + }
> >> +
> >> + rcmm->enabled = true;
> >> +
> >> + return 0;
> >> +}
> >> +EXPORT_SYMBOL_GPL(rcar_cmm_enable);
> >> +
> >> +/**
> >> + * rcar_cmm_disable() - disable the CMM unit
> >> + *
> >> + * Disable the CMM unit by stopping the parent clock.
> >> + *
> >> + * @pdev: The platform device associated with the CMM instance
> >
> > Parameters usually go before the description test.
>
> Indeed, sorry about this.
>
> >> + */
> >> +void rcar_cmm_disable(struct platform_device *pdev)
> >> +{
> >> + struct rcar_cmm *rcmm = platform_get_drvdata(pdev);
> >> +
> >> + rcar_cmm_write(rcmm, CM2_LUT_CTRL, 0);
> >> +
> >> + clk_disable_unprepare(rcmm->clk);
> >> +
> >> + rcmm->lut.restore = false;
> >> + rcmm->lut.running = false;
> >> + rcmm->lut.size = 0;
> >> +
> >> + rcmm->enabled = false;
> >> +}
> >> +EXPORT_SYMBOL_GPL(rcar_cmm_disable);
> >> +
> >> +#ifdef CONFIG_PM_SLEEP
> >> +static int rcar_cmm_pm_suspend(struct device *dev)
> >> +{
> >> + struct rcar_cmm *rcmm = dev_get_drvdata(dev);
> >> + unsigned int i;
> >> +
> >> + if (!(rcmm->lut.running || rcmm->lut.restore))
> >
> > Do you need the second part of this condition ? If a cached copy of the
> > LUT data has been stored but not applied yet because the CMM is
> > disabled, why would you need to overwrite that cached copy with values
> > from the hardware ?
>
> You are right, the second part of the condition is not needed (if not
> wrong).
>
> > I think you should have a first check on rcmm->enabled :
> >
> > if (!rcmm->enabled)
> > return 0;
> >
> > as in that case you can't access the hardware. Then, you can save the
> > LUT content only if it's running :
> >
> > if (rcmm->lut.running) {
> > /* Save the content */
> > ...
> > rcar_cmm_write(rcmm, CM2_LUT_CTRL, 0);
> > }
> >
> > I wouldn't clear rcmm->lut.running here, as from a software point of
> > view I think we still want to record that it's enabled. There's also no
> > need to touch the restore flag, see below.
> >
> >> + return 0;
> >> +
> >> + /* Save the LUT table entries in the scratch buffer table. */
> >
> > Should we call this a cache instead of a scratch buffer ?
> >
> >> + for (i = 0; i < rcmm->lut.size; ++i) {
> >> + int entry = rcar_cmm_read(rcmm, CM2_LUT_TBLA(i));
> >> + struct drm_color_lut *lut = &rcmm->lut.table[i];
> >> +
> >> + lut->blue = entry & 0xff;
> >> + lut->green = (entry >> 8) & 0xff;
> >> + lut->red = (entry >> 16) & 0xff;
> >> + }
> >> +
> >> + rcmm->lut.restore = true;
> >> + rcmm->lut.running = false;
> >> +
> >> + rcar_cmm_write(rcmm, CM2_LUT_CTRL, 0);
> >> +
> >> + return 0;
> >> +}
> >> +
> >> +static int rcar_cmm_pm_resume(struct device *dev)
> >> +{
> >> + struct rcar_cmm *rcmm = dev_get_drvdata(dev);
> >> +
> >> + if (!rcmm->lut.restore)
> >> + return 0;
> >> +
> >> + /* Program the LUT entries saved at suspend time. */
> >> + rcar_cmm_write(rcmm, CM2_LUT_CTRL, CM2_LUT_CTRL_EN);
> >> + rcar_cmm_lut_write(rcmm, rcmm->lut.size, rcmm->lut.table);
> >> + rcmm->lut.running = true;
> >> + rcmm->lut.restore = false;
> >
> > To be kept in sync with the modifications proposed above,
> >
> >
> > if (!rcmm->enabled)
> > return 0;
> >
> > if (rcmm->lut.running) {
> > /* Program the LUT entries saved at suspend time. */
> > rcar_cmm_write(rcmm, CM2_LUT_CTRL, CM2_LUT_CTRL_EN);
> > rcar_cmm_lut_write(rcmm, rcmm->lut.size, rcmm->lut.table);
> > }
> >
> > I think you can remove the restore field completely, as it's the only
> > used in rcar_cmm_enable(), and there you can check rcmm->lut.running
> > instead if you set rcmm->lut.running to true in rcar_cmm_setup() when
> > the CMM is disabled and the config requests the LUT to be enabled. The
> > overall logic should become simpler, with less corner cases.
>
> Thanks, very good suggestion, I probably don't need any restore flag
> at all. I'll see how it looks like, thanks.
>
> >> +
> >> + return 0;
> >> +}
> >> +#endif
> >> +
> >> +static const struct dev_pm_ops rcar_cmm_pm_ops = {
> >> + SET_SYSTEM_SLEEP_PM_OPS(rcar_cmm_pm_suspend, rcar_cmm_pm_resume)
> >> +};
> >> +
> >> +static int rcar_cmm_probe(struct platform_device *pdev)
> >> +{
> >> + struct rcar_cmm *rcmm;
> >> + struct resource *res;
> >> + resource_size_t size;
> >> +
> >> + rcmm = devm_kzalloc(&pdev->dev, sizeof(*rcmm), GFP_KERNEL);
> >> + if (!rcmm)
> >> + return -ENOMEM;
> >> +
> >> + platform_set_drvdata(pdev, rcmm);
> >> +
> >> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> >> + size = resource_size(res);
> >> + if (!devm_request_mem_region(&pdev->dev, res->start, size,
> >> + dev_name(&pdev->dev))) {
> >> + dev_err(&pdev->dev,
> >> + "can't request region for resource %pR\n", res);
> >> + return -EBUSY;
> >> + }
> >> +
> >> + rcmm->base = devm_ioremap_nocache(&pdev->dev, res->start, size);
> >> + if (IS_ERR(rcmm->base))
> >> + return PTR_ERR(rcmm->base);
> >
> > I still think you can use devm_ioremap_resource(). I agree it doesn't
> > explicitly request an uncached mapping, but I think the magic happens
> > behind the scene with the IO accessors to ensure it will work fine.
>
> Probably, but does using the _nocache version hurt somehow ?
Probably not, but it's more code :-) kmalloc + memset doesn't hurt
either, but kzalloc is still preferred.
> >> +
> >> + rcmm->clk = devm_clk_get(&pdev->dev, NULL);
> >> + if (IS_ERR(rcmm->clk)) {
> >> + dev_err(&pdev->dev, "Failed to get CMM clock");
> >> + return PTR_ERR(rcmm->clk);
> >> + }
> >> +
> >> + return 0;
> >> +}
> >> +
> >> +static const struct of_device_id rcar_cmm_of_table[] = {
> >> + { .compatible = "renesas,rcar-gen3-cmm", },
> >> + { .compatible = "renesas,rcar-gen2-cmm", },
> >> + { },
> >> +};
> >> +
> >> +MODULE_DEVICE_TABLE(of, rcar_cmm_of_table);
> >> +
> >> +static struct platform_driver rcar_cmm_platform_driver = {
> >> + .probe = rcar_cmm_probe,
> >> + .driver = {
> >> + .name = "rcar-cmm",
> >> + .pm = &rcar_cmm_pm_ops,
> >> + .of_match_table = rcar_cmm_of_table,
> >> + },
> >> +};
> >> +
> >> +module_platform_driver(rcar_cmm_platform_driver);
> >> +
> >> +MODULE_AUTHOR("Jacopo Mondi <jacopo+renesas@xxxxxxxxxx>");
> >> +MODULE_DESCRIPTION("Renesas R-Car CMM Driver");
> >> +MODULE_LICENSE("GPL v2");
> >> diff --git a/drivers/gpu/drm/rcar-du/rcar_cmm.h b/drivers/gpu/drm/rcar-du/rcar_cmm.h
> >> new file mode 100644
> >> index 000000000000..8744e72f32cd
> >> --- /dev/null
> >> +++ b/drivers/gpu/drm/rcar-du/rcar_cmm.h
> >> @@ -0,0 +1,38 @@
> >> +/* SPDX-License-Identifier: GPL-2.0+ */
> >> +/*
> >> + * rcar_cmm.h -- R-Car Display Unit Color Management Module
> >> + *
> >> + * Copyright (C) 2019 Jacopo Mondi <jacopo+renesas@xxxxxxxxxx>
> >> + */
> >> +
> >> +#ifndef __RCAR_CMM_H__
> >> +#define __RCAR_CMM_H__
> >> +
> >> +#define CMM_GAMMA_LUT_SIZE 256
> >> +
> >> +struct platform_device;
> >> +struct drm_color_lut;
> >
> > Could you please sort the forward declarations alphabetically ?
> >
> >> +
> >> +/**
> >> + * struct rcar_cmm_config - CMM configuration
> >> + *
> >> + * @lut: 1D-LUT configuration
> >> + * @lut.enable: 1D-LUT enable flag
> >> + * @lut.table: 1D-LUT table entries.
> >
> > s/\.$//
> >
> >> + * @lut.size 1D-LUT number of entries. Max is 256.
> >
> > "Number of 1D-LUT entries (max 256)"
> >
>
> And missing : after @lut.size
>
> >> + */
> >> +struct rcar_cmm_config {
> >> + struct {
> >> + bool enable;
> >> + struct drm_color_lut *table;
> >> + unsigned int size;
> >> + } lut;
> >> +};
> >> +
> >> +int rcar_cmm_enable(struct platform_device *pdev);
> >> +void rcar_cmm_disable(struct platform_device *pdev);
> >> +
> >> +int rcar_cmm_setup(struct platform_device *pdev,
> >> + const struct rcar_cmm_config *config);
> >> +
> >> +#endif /* __RCAR_CMM_H__ */
--
Regards,
Laurent Pinchart