by Marijn Suijten

[permalink] [raw]

Subject: Re: [PATCH 3/3] drm: panel: Add LG sw43408 panel driver

On 2024-04-01 22:11:48, Dmitry Baryshkov wrote:
> On Mon, 1 Apr 2024 at 13:29, Marijn Suijten
> <[email protected]> wrote:
> >
> > On 2024-03-30 16:37:08, Dmitry Baryshkov wrote:
> > > On Sat, 30 Mar 2024 at 12:27, Marijn Suijten
> > > <[email protected]> wrote:
> > > >
> > > > On 2024-03-30 05:59:30, Dmitry Baryshkov wrote:
> > > > > From: Sumit Semwal <[email protected]>
> > > > >
> > > > > LG SW43408 is 1080x2160, 4-lane MIPI-DSI panel, used in some Pixel3
> > > > > phones.
> > > > >
> > > > > Whatever init sequence we have for this panel isn't capable of
> > > > > initialising it completely, toggling the reset gpio ever causes the
> > > > > panel to die. Until this is resolved we avoid resetting the panel. The
> > > >
> > > > Are you sure it is avoided? This patch seems to be toggling reset_gpio in
> > > > sw43408_prepare()?
> > > >
> > > > > disable/unprepare functions only put the panel to sleep mode and
> > > > > disable the backlight.
> > > > >
> > > > > Signed-off-by: Sumit Semwal <[email protected]>
> > > > > [vinod: Add DSC support]
> > > > > Signed-off-by: Vinod Koul <[email protected]>
> > > > > [caleb: cleanup and support turning off the panel]
> > > > > Signed-off-by: Caleb Connolly <[email protected]>
> > > > > [DB: partially rewrote the driver and fixed DSC programming]
> > > > > Signed-off-by: Dmitry Baryshkov <[email protected]>
> > > > > ---
> > > > > MAINTAINERS | 8 +
> > > > > drivers/gpu/drm/panel/Kconfig | 11 ++
> > > > > drivers/gpu/drm/panel/Makefile | 1 +
> > > > > drivers/gpu/drm/panel/panel-lg-sw43408.c | 322 +++++++++++++++++++++++++++++++
> > > > > 4 files changed, 342 insertions(+)
> > > > >
> > > > > diff --git a/MAINTAINERS b/MAINTAINERS
> > > > > index 4b511a55101c..f4cf7ee97376 100644
> > > > > --- a/MAINTAINERS
> > > > > +++ b/MAINTAINERS
> > > > > @@ -6755,6 +6755,14 @@ S: Maintained
> > > > > F: Documentation/devicetree/bindings/display/panel/jadard,jd9365da-h3.yaml
> > > > > F: drivers/gpu/drm/panel/panel-jadard-jd9365da-h3.c
> > > > >
> > > > > +DRM DRIVER FOR LG SW43408 PANELS
> > > > > +M: Sumit Semwal <[email protected]>
> > > > > +M: Caleb Connolly <[email protected]>
> > > > > +S: Maintained
> > > > > +T: git git://anongit.freedesktop.org/drm/drm-misc
> > > > > +F: Documentation/devicetree/bindings/display/panel/lg,sw43408.yaml
> > > > > +F: drivers/gpu/drm/panel/panel-lg-sw43408.c
> > > > > +
> > > > > DRM DRIVER FOR LOGICVC DISPLAY CONTROLLER
> > > > > M: Paul Kocialkowski <[email protected]>
> > > > > S: Supported
> > > > > diff --git a/drivers/gpu/drm/panel/Kconfig b/drivers/gpu/drm/panel/Kconfig
> > > > > index d037b3b8b999..f94c702735cb 100644
> > > > > --- a/drivers/gpu/drm/panel/Kconfig
> > > > > +++ b/drivers/gpu/drm/panel/Kconfig
> > > > > @@ -335,6 +335,17 @@ config DRM_PANEL_LG_LG4573
> > > > > Say Y here if you want to enable support for LG4573 RGB panel.
> > > > > To compile this driver as a module, choose M here.
> > > > >
> > > > > +config DRM_PANEL_LG_SW43408
> > > > > + tristate "LG SW43408 panel"
> > > > > + depends on OF
> > > > > + depends on DRM_MIPI_DSI
> > > > > + depends on BACKLIGHT_CLASS_DEVICE
> > > > > + help
> > > > > + Say Y here if you want to enable support for LG sw43408 panel.
> > > > > + The panel has a 1080x2160 resolution and uses
> > > > > + 24 bit RGB per pixel. It provides a MIPI DSI interface to
> > > > > + the host and has a built-in LED backlight.
> > > > > +
> > > > > config DRM_PANEL_MAGNACHIP_D53E6EA8966
> > > > > tristate "Magnachip D53E6EA8966 DSI panel"
> > > > > depends on OF && SPI
> > > > > diff --git a/drivers/gpu/drm/panel/Makefile b/drivers/gpu/drm/panel/Makefile
> > > > > index f156d7fa0bcc..a75687d13caf 100644
> > > > > --- a/drivers/gpu/drm/panel/Makefile
> > > > > +++ b/drivers/gpu/drm/panel/Makefile
> > > > > @@ -34,6 +34,7 @@ obj-$(CONFIG_DRM_PANEL_LEADTEK_LTK050H3146W) += panel-leadtek-ltk050h3146w.o
> > > > > obj-$(CONFIG_DRM_PANEL_LEADTEK_LTK500HD1829) += panel-leadtek-ltk500hd1829.o
> > > > > obj-$(CONFIG_DRM_PANEL_LG_LB035Q02) += panel-lg-lb035q02.o
> > > > > obj-$(CONFIG_DRM_PANEL_LG_LG4573) += panel-lg-lg4573.o
> > > > > +obj-$(CONFIG_DRM_PANEL_LG_SW43408) += panel-lg-sw43408.o
> > > > > obj-$(CONFIG_DRM_PANEL_MAGNACHIP_D53E6EA8966) += panel-magnachip-d53e6ea8966.o
> > > > > obj-$(CONFIG_DRM_PANEL_NEC_NL8048HL11) += panel-nec-nl8048hl11.o
> > > > > obj-$(CONFIG_DRM_PANEL_NEWVISION_NV3051D) += panel-newvision-nv3051d.o
> > > > > diff --git a/drivers/gpu/drm/panel/panel-lg-sw43408.c b/drivers/gpu/drm/panel/panel-lg-sw43408.c
> > > > > new file mode 100644
> > > > > index 000000000000..365d25e14d54
> > > > > --- /dev/null
> > > > > +++ b/drivers/gpu/drm/panel/panel-lg-sw43408.c
> > > > > @@ -0,0 +1,322 @@
> > > > > +// SPDX-License-Identifier: GPL-2.0+
> > > > > +/*
> > > > > + * Copyright (C) 2019-2024 Linaro Ltd
> > > > > + * Author: Sumit Semwal <[email protected]>
> > > > > + * Dmitry Baryshkov <[email protected]>
> > > > > + */
> > > > > +
> > > > > +#include <linux/backlight.h>
> > > > > +#include <linux/delay.h>
> > > > > +#include <linux/gpio/consumer.h>
> > > > > +#include <linux/module.h>
> > > > > +#include <linux/of.h>
> > > > > +#include <linux/regulator/consumer.h>
> > > > > +
> > > > > +#include <video/mipi_display.h>
> > > > > +
> > > > > +#include <drm/drm_mipi_dsi.h>
> > > > > +#include <drm/drm_panel.h>
> > > > > +#include <drm/drm_probe_helper.h>
> > > > > +#include <drm/display/drm_dsc.h>
> > > > > +#include <drm/display/drm_dsc_helper.h>
> > > > > +
> > > > > +#define NUM_SUPPLIES 2
> > > > > +
> > > > > +struct sw43408_panel {
> > > > > + struct drm_panel base;
> > > > > + struct mipi_dsi_device *link;
> > > > > +
> > > > > + const struct drm_display_mode *mode;
> > > > > +
> > > > > + struct regulator_bulk_data supplies[NUM_SUPPLIES];
> > > > > +
> > > > > + struct gpio_desc *reset_gpio;
> > > > > +};
> > > > > +
> > > > > +static inline struct sw43408_panel *to_panel_info(struct drm_panel *panel)
> > > > > +{
> > > > > + return container_of(panel, struct sw43408_panel, base);
> > > > > +}
> > > > > +
> > > > > +static int sw43408_unprepare(struct drm_panel *panel)
> > > > > +{
> > > > > + struct sw43408_panel *ctx = to_panel_info(panel);
> > > > > + int ret;
> > > > > +
> > > > > + ret = mipi_dsi_dcs_set_display_off(ctx->link);
> > > > > + if (ret < 0)
> > > > > + dev_err(panel->dev, "set_display_off cmd failed ret = %d\n", ret);
> > > > > +
> > > > > + ret = mipi_dsi_dcs_enter_sleep_mode(ctx->link);
> > > > > + if (ret < 0)
> > > > > + dev_err(panel->dev, "enter_sleep cmd failed ret = %d\n", ret);
> > > > > +
> > > > > + msleep(100);
> > > > > +
> > > > > + gpiod_set_value(ctx->reset_gpio, 1);
> > > > > +
> > > > > + return regulator_bulk_disable(ARRAY_SIZE(ctx->supplies), ctx->supplies);
> > > > > +}
> > > > > +
> > > > > +static int sw43408_program(struct drm_panel *panel)
> > > > > +{
> > > > > + struct sw43408_panel *ctx = to_panel_info(panel);
> > > > > + struct drm_dsc_picture_parameter_set pps;
> > > > > + u8 dsc_en = 0x11;
> > > >
> > > > Yeah, this is completely strange. Bit 0, 0x1, is to enable DSC which is
> > > > normal. 0x10 however, which is bit 4, selects PPS table 2. Do you ever set
> > > > pps_identifier in struct drm_dsc_picture_parameter_set to 2? Or is the table
> > > > that you send below bogus and/or not used? Maybe the Driver IC on the other
> > > > end of the DSI link has a default PPS table with identifier 2 that works out of
> > > > the box?
> > >
> > > Note, MIPI standard also requires two bytes argument. I suspect that
> > > LG didn't fully follow the standard here.
> >
> > Have you read this command from downstream DTS, or have you tried sending 2
> > bytes and seen the panel breaking? The second byte is marked as reserved and
> > should be equal to 0; if the Driver IC is okay with sending either 1 or 2 bytes
> > I'd strive to stick with the defined length of 2 bytes for this DCS.
> >
> > Have you played around with the PPS table? What if you change
> > drm_dsc_picture_paremeter_set::pps_identifier to the second table, will the
> > panel stop working as expected again? This could indicate that the PPS that is
> > sent is incorrect (even though the information in the original DSC config was
> > enough to set up the DPU and DSI correctly).
> >
> > According to the DSI spec it is allowed to have a pre-stored/pre-programmed
> > PPS table, which could be used here making the current call to
> > mipi_dsi_picture_parameter_set() useless and "confusing"?
>
> Ok, some short summary of my tests.
>
> Skipping PPS doesn't work at all, so there is no default.
>
> Adding a second zero byte doesn't seem to change anything. Dropping
> the 0x1 bit ('enable') doesn't seem to change anything.
>
> If I send COMPRESSION_MODE before sending the PPS, various combinations work.
> If I send COMPRESSION_MODE after sending the PPS, the follow combos work:
>
> pps_identifier = 0x0, COMPRESSION_MODE = 0x11
> pps_identifier = 0x1, COMPRESSION_MODE = 0x21

Thanks, this must really be an off-by-one table identifier. I presume you've
tested pps_identifier=0x2 with COMPRESSION_MODE=0x31, and that there are only 2
tables and not 3 or 4?

From this we can also assume that sending a new PPS will automatically switch
the compression mode to the pps_identifier in that PPS, COMPRESSION_MODE doesn't
seem to affect it when sent too early.

> > > Basically that's the reason why I went for the _raw function instead
> > > of adding PPS and codec arguments to the existing function.
> > >
> > > >
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, MIPI_DCS_SET_GAMMA_CURVE, 0x02);
> > > > > +
> > > > > + mipi_dsi_dcs_set_tear_on(ctx->link, MIPI_DSI_DCS_TEAR_MODE_VBLANK);
> > > > > +
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0x53, 0x0c, 0x30);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0x55, 0x00, 0x70, 0xdf, 0x00, 0x70, 0xdf);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xf7, 0x01, 0x49, 0x0c);
> > > > > +
> > > > > + mipi_dsi_dcs_exit_sleep_mode(ctx->link);
> > > > > +
> > > > > + msleep(135);
> > > > > +
> > > > > + mipi_dsi_compression_mode_raw(ctx->link, &dsc_en, 1);
> > > >
> > > > Even though I think we should change this function to describe the known
> > > > bit layout of command 0x7 per the VESA DSI spec, for now replace 1 with
> > > > sizeof(dsc_en)?
> > >
> > > If dsc_en were an array, it would have been a proper thing. Maybe I
> > > should change it to the array to remove confusion.
> >
> > It should work even with a single byte, just to clarify to readers that the 3rd
> > argument is the byte-size of the input.
> >
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xb0, 0xac);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xe5,
> > > > > + 0x00, 0x3a, 0x00, 0x3a, 0x00, 0x0e, 0x10);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xb5,
> > > > > + 0x75, 0x60, 0x2d, 0x5d, 0x80, 0x00, 0x0a, 0x0b,
> > > > > + 0x00, 0x05, 0x0b, 0x00, 0x80, 0x0d, 0x0e, 0x40,
> > > > > + 0x00, 0x0c, 0x00, 0x16, 0x00, 0xb8, 0x00, 0x80,
> > > > > + 0x0d, 0x0e, 0x40, 0x00, 0x0c, 0x00, 0x16, 0x00,
> > > > > + 0xb8, 0x00, 0x81, 0x00, 0x03, 0x03, 0x03, 0x01,
> > > > > + 0x01);
> > > > > + msleep(85);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xcd,
> > > > > + 0x00, 0x00, 0x00, 0x19, 0x19, 0x19, 0x19, 0x19,
> > > > > + 0x19, 0x19, 0x19, 0x19, 0x19, 0x19, 0x19, 0x19,
> > > > > + 0x16, 0x16);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xcb, 0x80, 0x5c, 0x07, 0x03, 0x28);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xc0, 0x02, 0x02, 0x0f);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0x55, 0x04, 0x61, 0xdb, 0x04, 0x70, 0xdb);
> > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xb0, 0xca);
> > > > > +
> > > > > + mipi_dsi_dcs_set_display_on(ctx->link);
> > > >
> > > > Any specific reason to not have the (un)blanking sequence in the enable/disable
> > > > callbacks and leaving display configuration in (un)prepare?
> > >
> > > We are back to the question on when it's fine to send the commands. I
> > > think the current agreement is to send everything in the
> > > prepare/unprepare, because of some strange hosts.
> >
> > For my panel drivers I'm sticking with having `post-on` commands (from
> > downstream) in `enable/disable`, which is typically only `set_display_on`. In
> > hopes of proposing a `prepare_atomic()` some time to allow mode selection.
> >
> > In a short test on recent -next I am once again allowed to send DSI commands in
> > both .disable and .unprepare, making both functions a "clean" inverse of .enable
> > and .prepare respectively.
>
> The world isn't limited to the MSM hosts.

If I'm not mistaken this was an ordering issue in the drm_bridge implementation.

But you are right that some hosts might not be all too happy with sending
commands (like unblanking?) after the cmd/video stream started, and before the
stream stops. Which, as far as I know, are what .enable and .disable do. On
the other hand, I was under the impression that this split mainly existed to do
all the heavy/required lifting up-front, and only unblank when there's a video
signal to combat any possible observed corruption?

In the end I'm just curious if there's a specific reason - that I need
to take into account when resending all my panel patches - to /not/
use .enable/.disable?

- Marijn

> > > > > + msleep(50);
> > > > > +
> > > > > + ctx->link->mode_flags &= ~MIPI_DSI_MODE_LPM;
> > > > > +
> > > > > + drm_dsc_pps_payload_pack(&pps, ctx->link->dsc);
> > > > > + mipi_dsi_picture_parameter_set(ctx->link, &pps);
> > > >
> > > > I'm always surprised why this is sent _after_ turning the display on (unblanking
> > > > it). Wouldn't that cause unnecessary corruption?
> > >
> > > No idea. I followed the dowsntream command sequences here. Most likely
> > > the panel is not fully on until it receives the full frame to be
> > > displayed.
> >
> > According to the DSI spec a PPS update is allowed to happen every frame, and
> > (for cmdmode panels) will take effect after the next TE trigger. Unsure if a TE
> > event happens before the first frame, otherwise this may start taking effect
> > on the second frame onwards only.
> >
> > If there's no corruption on the first frame there might be a pre-programmed PPS
> > table in slot 2, supporting the theory above.
>
>
>
> --
> With best wishes
> Dmitry

2024-04-03 00:51:51

by Dmitry Baryshkov

[permalink] [raw]

Subject: Re: [PATCH 3/3] drm: panel: Add LG sw43408 panel driver

On Tue, 2 Apr 2024 at 23:57, Marijn Suijten
<[email protected]> wrote:
>
> On 2024-04-01 22:11:48, Dmitry Baryshkov wrote:
> > On Mon, 1 Apr 2024 at 13:29, Marijn Suijten
> > <[email protected]> wrote:
> > >
> > > On 2024-03-30 16:37:08, Dmitry Baryshkov wrote:
> > > > On Sat, 30 Mar 2024 at 12:27, Marijn Suijten
> > > > <[email protected]> wrote:
> > > > >
> > > > > On 2024-03-30 05:59:30, Dmitry Baryshkov wrote:
> > > > > > From: Sumit Semwal <[email protected]>
> > > > > >
> > > > > > LG SW43408 is 1080x2160, 4-lane MIPI-DSI panel, used in some Pixel3
> > > > > > phones.
> > > > > >
> > > > > > Whatever init sequence we have for this panel isn't capable of
> > > > > > initialising it completely, toggling the reset gpio ever causes the
> > > > > > panel to die. Until this is resolved we avoid resetting the panel. The
> > > > >
> > > > > Are you sure it is avoided? This patch seems to be toggling reset_gpio in
> > > > > sw43408_prepare()?
> > > > >
> > > > > > disable/unprepare functions only put the panel to sleep mode and
> > > > > > disable the backlight.
> > > > > >
> > > > > > Signed-off-by: Sumit Semwal <[email protected]>
> > > > > > [vinod: Add DSC support]
> > > > > > Signed-off-by: Vinod Koul <[email protected]>
> > > > > > [caleb: cleanup and support turning off the panel]
> > > > > > Signed-off-by: Caleb Connolly <[email protected]>
> > > > > > [DB: partially rewrote the driver and fixed DSC programming]
> > > > > > Signed-off-by: Dmitry Baryshkov <[email protected]>
> > > > > > ---
> > > > > > MAINTAINERS | 8 +
> > > > > > drivers/gpu/drm/panel/Kconfig | 11 ++
> > > > > > drivers/gpu/drm/panel/Makefile | 1 +
> > > > > > drivers/gpu/drm/panel/panel-lg-sw43408.c | 322 +++++++++++++++++++++++++++++++
> > > > > > 4 files changed, 342 insertions(+)
> > > > > >
> > > > > > diff --git a/MAINTAINERS b/MAINTAINERS
> > > > > > index 4b511a55101c..f4cf7ee97376 100644
> > > > > > --- a/MAINTAINERS
> > > > > > +++ b/MAINTAINERS
> > > > > > @@ -6755,6 +6755,14 @@ S: Maintained
> > > > > > F: Documentation/devicetree/bindings/display/panel/jadard,jd9365da-h3.yaml
> > > > > > F: drivers/gpu/drm/panel/panel-jadard-jd9365da-h3.c
> > > > > >
> > > > > > +DRM DRIVER FOR LG SW43408 PANELS
> > > > > > +M: Sumit Semwal <[email protected]>
> > > > > > +M: Caleb Connolly <[email protected]>
> > > > > > +S: Maintained
> > > > > > +T: git git://anongit.freedesktop.org/drm/drm-misc
> > > > > > +F: Documentation/devicetree/bindings/display/panel/lg,sw43408.yaml
> > > > > > +F: drivers/gpu/drm/panel/panel-lg-sw43408.c
> > > > > > +
> > > > > > DRM DRIVER FOR LOGICVC DISPLAY CONTROLLER
> > > > > > M: Paul Kocialkowski <[email protected]>
> > > > > > S: Supported
> > > > > > diff --git a/drivers/gpu/drm/panel/Kconfig b/drivers/gpu/drm/panel/Kconfig
> > > > > > index d037b3b8b999..f94c702735cb 100644
> > > > > > --- a/drivers/gpu/drm/panel/Kconfig
> > > > > > +++ b/drivers/gpu/drm/panel/Kconfig
> > > > > > @@ -335,6 +335,17 @@ config DRM_PANEL_LG_LG4573
> > > > > > Say Y here if you want to enable support for LG4573 RGB panel.
> > > > > > To compile this driver as a module, choose M here.
> > > > > >
> > > > > > +config DRM_PANEL_LG_SW43408
> > > > > > + tristate "LG SW43408 panel"
> > > > > > + depends on OF
> > > > > > + depends on DRM_MIPI_DSI
> > > > > > + depends on BACKLIGHT_CLASS_DEVICE
> > > > > > + help
> > > > > > + Say Y here if you want to enable support for LG sw43408 panel.
> > > > > > + The panel has a 1080x2160 resolution and uses
> > > > > > + 24 bit RGB per pixel. It provides a MIPI DSI interface to
> > > > > > + the host and has a built-in LED backlight.
> > > > > > +
> > > > > > config DRM_PANEL_MAGNACHIP_D53E6EA8966
> > > > > > tristate "Magnachip D53E6EA8966 DSI panel"
> > > > > > depends on OF && SPI
> > > > > > diff --git a/drivers/gpu/drm/panel/Makefile b/drivers/gpu/drm/panel/Makefile
> > > > > > index f156d7fa0bcc..a75687d13caf 100644
> > > > > > --- a/drivers/gpu/drm/panel/Makefile
> > > > > > +++ b/drivers/gpu/drm/panel/Makefile
> > > > > > @@ -34,6 +34,7 @@ obj-$(CONFIG_DRM_PANEL_LEADTEK_LTK050H3146W) += panel-leadtek-ltk050h3146w.o
> > > > > > obj-$(CONFIG_DRM_PANEL_LEADTEK_LTK500HD1829) += panel-leadtek-ltk500hd1829.o
> > > > > > obj-$(CONFIG_DRM_PANEL_LG_LB035Q02) += panel-lg-lb035q02.o
> > > > > > obj-$(CONFIG_DRM_PANEL_LG_LG4573) += panel-lg-lg4573.o
> > > > > > +obj-$(CONFIG_DRM_PANEL_LG_SW43408) += panel-lg-sw43408.o
> > > > > > obj-$(CONFIG_DRM_PANEL_MAGNACHIP_D53E6EA8966) += panel-magnachip-d53e6ea8966.o
> > > > > > obj-$(CONFIG_DRM_PANEL_NEC_NL8048HL11) += panel-nec-nl8048hl11.o
> > > > > > obj-$(CONFIG_DRM_PANEL_NEWVISION_NV3051D) += panel-newvision-nv3051d.o
> > > > > > diff --git a/drivers/gpu/drm/panel/panel-lg-sw43408.c b/drivers/gpu/drm/panel/panel-lg-sw43408.c
> > > > > > new file mode 100644
> > > > > > index 000000000000..365d25e14d54
> > > > > > --- /dev/null
> > > > > > +++ b/drivers/gpu/drm/panel/panel-lg-sw43408.c
> > > > > > @@ -0,0 +1,322 @@
> > > > > > +// SPDX-License-Identifier: GPL-2.0+
> > > > > > +/*
> > > > > > + * Copyright (C) 2019-2024 Linaro Ltd
> > > > > > + * Author: Sumit Semwal <[email protected]>
> > > > > > + * Dmitry Baryshkov <[email protected]>
> > > > > > + */
> > > > > > +
> > > > > > +#include <linux/backlight.h>
> > > > > > +#include <linux/delay.h>
> > > > > > +#include <linux/gpio/consumer.h>
> > > > > > +#include <linux/module.h>
> > > > > > +#include <linux/of.h>
> > > > > > +#include <linux/regulator/consumer.h>
> > > > > > +
> > > > > > +#include <video/mipi_display.h>
> > > > > > +
> > > > > > +#include <drm/drm_mipi_dsi.h>
> > > > > > +#include <drm/drm_panel.h>
> > > > > > +#include <drm/drm_probe_helper.h>
> > > > > > +#include <drm/display/drm_dsc.h>
> > > > > > +#include <drm/display/drm_dsc_helper.h>
> > > > > > +
> > > > > > +#define NUM_SUPPLIES 2
> > > > > > +
> > > > > > +struct sw43408_panel {
> > > > > > + struct drm_panel base;
> > > > > > + struct mipi_dsi_device *link;
> > > > > > +
> > > > > > + const struct drm_display_mode *mode;
> > > > > > +
> > > > > > + struct regulator_bulk_data supplies[NUM_SUPPLIES];
> > > > > > +
> > > > > > + struct gpio_desc *reset_gpio;
> > > > > > +};
> > > > > > +
> > > > > > +static inline struct sw43408_panel *to_panel_info(struct drm_panel *panel)
> > > > > > +{
> > > > > > + return container_of(panel, struct sw43408_panel, base);
> > > > > > +}
> > > > > > +
> > > > > > +static int sw43408_unprepare(struct drm_panel *panel)
> > > > > > +{
> > > > > > + struct sw43408_panel *ctx = to_panel_info(panel);
> > > > > > + int ret;
> > > > > > +
> > > > > > + ret = mipi_dsi_dcs_set_display_off(ctx->link);
> > > > > > + if (ret < 0)
> > > > > > + dev_err(panel->dev, "set_display_off cmd failed ret = %d\n", ret);
> > > > > > +
> > > > > > + ret = mipi_dsi_dcs_enter_sleep_mode(ctx->link);
> > > > > > + if (ret < 0)
> > > > > > + dev_err(panel->dev, "enter_sleep cmd failed ret = %d\n", ret);
> > > > > > +
> > > > > > + msleep(100);
> > > > > > +
> > > > > > + gpiod_set_value(ctx->reset_gpio, 1);
> > > > > > +
> > > > > > + return regulator_bulk_disable(ARRAY_SIZE(ctx->supplies), ctx->supplies);
> > > > > > +}
> > > > > > +
> > > > > > +static int sw43408_program(struct drm_panel *panel)
> > > > > > +{
> > > > > > + struct sw43408_panel *ctx = to_panel_info(panel);
> > > > > > + struct drm_dsc_picture_parameter_set pps;
> > > > > > + u8 dsc_en = 0x11;
> > > > >
> > > > > Yeah, this is completely strange. Bit 0, 0x1, is to enable DSC which is
> > > > > normal. 0x10 however, which is bit 4, selects PPS table 2. Do you ever set
> > > > > pps_identifier in struct drm_dsc_picture_parameter_set to 2? Or is the table
> > > > > that you send below bogus and/or not used? Maybe the Driver IC on the other
> > > > > end of the DSI link has a default PPS table with identifier 2 that works out of
> > > > > the box?
> > > >
> > > > Note, MIPI standard also requires two bytes argument. I suspect that
> > > > LG didn't fully follow the standard here.
> > >
> > > Have you read this command from downstream DTS, or have you tried sending 2
> > > bytes and seen the panel breaking? The second byte is marked as reserved and
> > > should be equal to 0; if the Driver IC is okay with sending either 1 or 2 bytes
> > > I'd strive to stick with the defined length of 2 bytes for this DCS.
> > >
> > > Have you played around with the PPS table? What if you change
> > > drm_dsc_picture_paremeter_set::pps_identifier to the second table, will the
> > > panel stop working as expected again? This could indicate that the PPS that is
> > > sent is incorrect (even though the information in the original DSC config was
> > > enough to set up the DPU and DSI correctly).
> > >
> > > According to the DSI spec it is allowed to have a pre-stored/pre-programmed
> > > PPS table, which could be used here making the current call to
> > > mipi_dsi_picture_parameter_set() useless and "confusing"?
> >
> > Ok, some short summary of my tests.
> >
> > Skipping PPS doesn't work at all, so there is no default.
> >
> > Adding a second zero byte doesn't seem to change anything. Dropping
> > the 0x1 bit ('enable') doesn't seem to change anything.
> >
> > If I send COMPRESSION_MODE before sending the PPS, various combinations work.
> > If I send COMPRESSION_MODE after sending the PPS, the follow combos work:
> >
> > pps_identifier = 0x0, COMPRESSION_MODE = 0x11
> > pps_identifier = 0x1, COMPRESSION_MODE = 0x21
>
> Thanks, this must really be an off-by-one table identifier. I presume you've
> tested pps_identifier=0x2 with COMPRESSION_MODE=0x31, and that there are only 2
> tables and not 3 or 4?

I was not able to get either pps_identifier = 0x2 or
COMPRESSION_MODE=0x31 to work.

>
> From this we can also assume that sending a new PPS will automatically switch
> the compression mode to the pps_identifier in that PPS, COMPRESSION_MODE doesn't
> seem to affect it when sent too early.

No. Omitting COMPRESSION_MODE packet breaks the display. Actually,
this was one of the issues: we were sending an incorrect packet. So
both are required.

>
> > > > Basically that's the reason why I went for the _raw function instead
> > > > of adding PPS and codec arguments to the existing function.
> > > >
> > > > >
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, MIPI_DCS_SET_GAMMA_CURVE, 0x02);
> > > > > > +
> > > > > > + mipi_dsi_dcs_set_tear_on(ctx->link, MIPI_DSI_DCS_TEAR_MODE_VBLANK);
> > > > > > +
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0x53, 0x0c, 0x30);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0x55, 0x00, 0x70, 0xdf, 0x00, 0x70, 0xdf);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xf7, 0x01, 0x49, 0x0c);
> > > > > > +
> > > > > > + mipi_dsi_dcs_exit_sleep_mode(ctx->link);
> > > > > > +
> > > > > > + msleep(135);
> > > > > > +
> > > > > > + mipi_dsi_compression_mode_raw(ctx->link, &dsc_en, 1);
> > > > >
> > > > > Even though I think we should change this function to describe the known
> > > > > bit layout of command 0x7 per the VESA DSI spec, for now replace 1 with
> > > > > sizeof(dsc_en)?
> > > >
> > > > If dsc_en were an array, it would have been a proper thing. Maybe I
> > > > should change it to the array to remove confusion.
> > >
> > > It should work even with a single byte, just to clarify to readers that the 3rd
> > > argument is the byte-size of the input.
> > >
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xb0, 0xac);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xe5,
> > > > > > + 0x00, 0x3a, 0x00, 0x3a, 0x00, 0x0e, 0x10);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xb5,
> > > > > > + 0x75, 0x60, 0x2d, 0x5d, 0x80, 0x00, 0x0a, 0x0b,
> > > > > > + 0x00, 0x05, 0x0b, 0x00, 0x80, 0x0d, 0x0e, 0x40,
> > > > > > + 0x00, 0x0c, 0x00, 0x16, 0x00, 0xb8, 0x00, 0x80,
> > > > > > + 0x0d, 0x0e, 0x40, 0x00, 0x0c, 0x00, 0x16, 0x00,
> > > > > > + 0xb8, 0x00, 0x81, 0x00, 0x03, 0x03, 0x03, 0x01,
> > > > > > + 0x01);
> > > > > > + msleep(85);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xcd,
> > > > > > + 0x00, 0x00, 0x00, 0x19, 0x19, 0x19, 0x19, 0x19,
> > > > > > + 0x19, 0x19, 0x19, 0x19, 0x19, 0x19, 0x19, 0x19,
> > > > > > + 0x16, 0x16);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xcb, 0x80, 0x5c, 0x07, 0x03, 0x28);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xc0, 0x02, 0x02, 0x0f);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0x55, 0x04, 0x61, 0xdb, 0x04, 0x70, 0xdb);
> > > > > > + mipi_dsi_dcs_write_seq(ctx->link, 0xb0, 0xca);
> > > > > > +
> > > > > > + mipi_dsi_dcs_set_display_on(ctx->link);
> > > > >
> > > > > Any specific reason to not have the (un)blanking sequence in the enable/disable
> > > > > callbacks and leaving display configuration in (un)prepare?
> > > >
> > > > We are back to the question on when it's fine to send the commands. I
> > > > think the current agreement is to send everything in the
> > > > prepare/unprepare, because of some strange hosts.
> > >
> > > For my panel drivers I'm sticking with having `post-on` commands (from
> > > downstream) in `enable/disable`, which is typically only `set_display_on`. In
> > > hopes of proposing a `prepare_atomic()` some time to allow mode selection.
> > >
> > > In a short test on recent -next I am once again allowed to send DSI commands in
> > > both .disable and .unprepare, making both functions a "clean" inverse of .enable
> > > and .prepare respectively.
> >
> > The world isn't limited to the MSM hosts.
>
> If I'm not mistaken this was an ordering issue in the drm_bridge implementation.

No. I think sunxi didn't support sending DSI commands after starting
the video stream.

>
> But you are right that some hosts might not be all too happy with sending
> commands (like unblanking?) after the cmd/video stream started, and before the
> stream stops. Which, as far as I know, are what .enable and .disable do. On
> the other hand, I was under the impression that this split mainly existed to do
> all the heavy/required lifting up-front, and only unblank when there's a video
> signal to combat any possible observed corruption?

In the ideal world that would be true. But we are not living in the
ideal world. I still have hopes to get back to the idea of reworling
this part of the DSI framework.

> In the end I'm just curious if there's a specific reason - that I need
> to take into account when resending all my panel patches - to /not/
> use .enable/.disable?

I have bookmarked this email as a main reference for the topic:
https://lore.kernel.org/dri-devel/CAPY8ntBrhYAmsraDqJGuTrSL6VjGXBAMVoN7xweV7E4qZv+v3Q@mail.gmail.com/

>
> - Marijn
>
> > > > > > + msleep(50);
> > > > > > +
> > > > > > + ctx->link->mode_flags &= ~MIPI_DSI_MODE_LPM;
> > > > > > +
> > > > > > + drm_dsc_pps_payload_pack(&pps, ctx->link->dsc);
> > > > > > + mipi_dsi_picture_parameter_set(ctx->link, &pps);
> > > > >
> > > > > I'm always surprised why this is sent _after_ turning the display on (unblanking
> > > > > it). Wouldn't that cause unnecessary corruption?
> > > >
> > > > No idea. I followed the dowsntream command sequences here. Most likely
> > > > the panel is not fully on until it receives the full frame to be
> > > > displayed.
> > >
> > > According to the DSI spec a PPS update is allowed to happen every frame, and
> > > (for cmdmode panels) will take effect after the next TE trigger. Unsure if a TE
> > > event happens before the first frame, otherwise this may start taking effect
> > > on the second frame onwards only.
> > >
> > > If there's no corruption on the first frame there might be a pre-programmed PPS
> > > table in slot 2, supporting the theory above.
> >
> >
> >
> > --
> > With best wishes
> > Dmitry

--
With best wishes
Dmitry