RE: [PATCH 0/7] drm/exynos: move hdmiphy related code to hdmiphy driver

[Inki Dae <inki.dae@samsung.com>]

> -----Original Message-----
> From: Rahul Sharma [mailto:r.sh.open@gmail.com]
> Sent: Monday, September 02, 2013 6:06 PM
> To: Inki Dae
> Cc: Rahul Sharma; linux-samsung-soc; dri-devel@lists.freedesktop.org;
> Kukjin Kim; sw0312.kim; Sean Paul; Lucas Stach; Tomasz Figa; Sylwester
> Nawrocki; sunil joshi
> Subject: Re: [PATCH 0/7] drm/exynos: move hdmiphy related code to hdmiphy
> driver
> 
> On 2 September 2013 12:52, Inki Dae <inki.dae@samsung.com> wrote:
> >
> >
> >> -----Original Message-----
> >> From: Rahul Sharma [mailto:r.sh.open@gmail.com]
> >> Sent: Monday, September 02, 2013 3:28 PM
> >> To: Inki Dae
> >> Cc: Rahul Sharma; linux-samsung-soc; dri-devel@lists.freedesktop.org;
> >> Kukjin Kim; sw0312.kim; Sean Paul; Lucas Stach; Tomasz Figa; Sylwester
> >> Nawrocki; sunil joshi
> >> Subject: Re: [PATCH 0/7] drm/exynos: move hdmiphy related code to
> hdmiphy
> >> driver
> >>
> >> On 2 September 2013 10:38, Inki Dae <inki.dae@samsung.com> wrote:
> >> > Hi Rahul,
> >> >
> >> >> -----Original Message-----
> >> >> From: linux-samsung-soc-owner@vger.kernel.org [mailto:linux-samsung-
> >> soc-
> >> >> owner@vger.kernel.org] On Behalf Of Rahul Sharma
> >> >> Sent: Friday, August 30, 2013 7:06 PM
> >> >> To: Inki Dae
> >> >> Cc: Rahul Sharma; linux-samsung-soc;
dri-devel@lists.freedesktop.org;
> >> >> Kukjin Kim; sw0312.kim; Sean Paul; Lucas Stach; Tomasz Figa;
> Sylwester
> >> >> Nawrocki; sunil joshi
> >> >> Subject: Re: [PATCH 0/7] drm/exynos: move hdmiphy related code to
> >> hdmiphy
> >> >> driver
> >> >>
> >> >> Thanks Mr. Dae,
> >> >>
> >> >> I have some points for discussion.
> >> >>
> >> >> On 30 August 2013 14:03, Inki Dae <inki.dae@samsung.com> wrote:
> >> >> > Hi Rahul.
> >> >> >
> >> >> > Thanks for your patch set.
> >> >> >
> >> >> > I had just quick review to all patch series. And I think we could
> >> fully
> >> >> hide
> >> >> > hdmiphy interfaces,
> >> >> > exynos_hdmiphy_enable/disable/check_mode/set_mode/conf_apply, from
> >> hdmi
> >> >> > driver.
> >> >> > That may be prototyped like below,
> >> >> >
> >> >> > at exynos_hdmi.h
> >> >> >
> >> >> > /* Define hdmiphy callbacks. */
> >> >> > struct exynos_hdmiphy_ops {
> >> >> >         void (*enable)(struct device *dev);
> >> >> >         void (*disable)(struct device *dev);
> >> >> >         int (*check_mode)(struct device *dev, struct
> drm_display_mode
> >> >> > *mode);
> >> >> >         int (*set_mode)(struct device *dev, struct
drm_display_mode
> >> > *mode);
> >> >> >         int (*apply)(struct device *dev);
> >> >> > };
> >> >> >
> >> >>
> >> >> Above looks fine to me. I will hide the ops as you suggested.
> >> >>
> >> >> >
> >> >> > at exynos_hdmi.c
> >> >> >
> >> >> > /*
> >> >> >   * Add a new structure for hdmi driver data.
> >> >> >   * type could be HDMI_TYPE13 or HDMI_TYPE14.
> >> >> >   * i2c_hdmiphy could be true or false: true means that current
> hdmi
> >> >> device
> >> >> > uses i2c
> >> >> >   * based hdmiphy device, otherwise platform device based one.
> >> >> >   */
> >> >> > struct hdmi_drv_data {
> >> >> >         unsigned int type;
> >> >> >         unsigned int i2c_hdmiphy;
> >> >> > };
> >> >> >
> >> >> > ...
> >> >> >
> >> >> > /* Add new members to hdmi context. */
> >> >> > struct hdmi_context {
> >> >> >         ...
> >> >> >         struct hdmi_drv_data *drv_data;
> >> >> >         struct hdmiphy_ops *ops;
> >> >> >         ...
> >> >> > };
> >> >> >
> >> >> >
> >> >> > /* Add hdmi device data according Exynos SoC. */
> >> >> > static struct hdmi_drv_data exynos4212_hdmi_drv_data = {
> >> >> >         .type = HDMI_TYPE14,
> >> >> >         .i2c_hdmiphy = true
> >> >> > };
> >> >> >
> >> >> > static struct hdmi_drv_data exynos5420_hdmi_drv_data = {
> >> >> >         .type = HDMI_TYPE14,
> >> >> >         .i2c_hdmiphy = false
> >> >> > };
> >> >> >
> >> >> >
> >> >> > static struct of_device_id hdmi_match_types[] = {
> >> >> >         {
> >> >> >                 .compatible = "samsung,exynos4212-hdmi",
> >> >> >                 .data           = (void
*)&exynos4212_hdmi_drv_data,
> >> >> >         }, {
> >> >> >         ...
> >> >> >
> >> >> >                 .compatible = "samsung,exynos5420-hdmi",
> >> >> >                 .data           = (void
*)&exynos5420_hdmi_drv_data,
> >> >> >         }, {
> >> >> >         }
> >> >> > };
> >> >> >
> >> >> > /* the below example function shows how hdmiphy interfaces can be
> >> hided
> >> >> from
> >> >> > hdmi driver. */
> >> >> > static void hdmi_mode_set(...)
> >> >> > {
> >> >> >         ...
> >> >> >         hdata->ops->set_mode(hdata->hdmiphy_dev, mode);
> >> >>
> >> >> This looks fine.
> >> >>
> >> >> > }
> >> >> >
> >> >> > static int hdmi_get_phy_device(struct hdmi_context *hdata)
> >> >> > {
> >> >> >         struct hdmi_drv_data *data = hdata->drv_data;
> >> >> >
> >> >> >         ...
> >> >> >         /* Register hdmiphy driver according to i2c_hdmiphy value.
> */
> >> >> >         ret = exynos_hdmiphy_driver_register(data->i2c_hdmiphy);
> >> >> >         ...
> >> >> >         /* Get hdmiphy driver ops according to i2c_hdmiphy value.
*/
> >> >> >         hdata->ops = exynos_hdmiphy_get_ops(data->i2c_hdmiphy);
> >> >> >         ...
> >> >> > }
> >> >> >
> >> >> >
> >> >> > at exynos_hdmiphy.c
> >> >> >
> >> >> > /* Define hdmiphy ops respectively. */
> >> >> > struct exynos_hdmiphy_ops hdmiphy_i2c_ops = {
> >> >> >         .enable = exynos_hdmiphy_i2c_enable,
> >> >> >         .disable = exynos_hdmiphy_i2c_disable,
> >> >> >         ...
> >> >> > };
> >> >> >
> >> >> > struct exynos_hdmiphy_ops hdmiphy_platdev_ops = {
> >> >> >         .enable = exynos_hdmiphy_platdev_enable,
> >> >> >         .disable = exynos_hdmiphy_platdev_disable,
> >> >> >         ...
> >> >> > };
> >> >>
> >> >> Actually, Ops for Hdmiphy I2c and platform devices are exactly
> >> >> same. We don't need 2 sets. Only difference is in static function to
> >> >> write configuration values to phy. Based on i2c_verify_client(hdata-
> >> >dev),
> >> >> we use i2c_master_send or writeb.
> >> >>
> >> >> >
> >> >> > struct exynos_hdmiphy_ops *exynos_hdmiphy_get_ops(unsigned int
> >> >> i2c_hdmiphy)
> >> >> > {
> >> >> >         /* Return hdmiphy ops appropriately according to
i2c_hdmiphy.
> >> */
> >> >> >         if (i2c_hdmiphy)
> >> >> >                 return &hdmiphy_i2c_ops;
> >> >> >
> >> >> >         return &hdmiphy_platdev_ops;
> >> >> > }
> >> >>
> >> >> We don't need i2c_hdmiphy flag from the hdmi driver. After probe,
> >> >> this information is available in hdmiphy driver itself.
> >> >>
> >> >> >
> >> >> > int exynos_hdmiphy_driver_register(unsigned int i2c_hdmiphy)
> >> >> > {
> >> >> >         ...
> >> >> >         /* Register hdmiphy driver appropriately according to
> >> > i2c_hdmiphy.
> >> >> > */
> >> >> >         if (i2c_hdmiphy) {
> >> >> >                 ret = i2c_add_driver(&hdmiphy_i2c_driver);
> >> >> >                 ...
> >> >> >         } else {
> >> >> >                 ret =
> >> > platform_driver_register(&hdmiphy_platform_driver);
> >> >> >                 ...
> >> >> >         }
> >> >> >
> >> >>
> >> >> Here i2c_hdmiphy flag helps in avoiding registering both i2c
> >> >> and platform drivers for phy. But is it a concern that we should
> >> >> not register 2 drivers on different buses for single hw device. I am
> >> >> still not clear on this.
> >> >>
> >> >> Otherwise we do not need to maintain i2c_hdmiphy flag.
> >> >>
> >> >> Secondly, we always have registration of i2c driver for ddc and
> >> >> hdmiphy driver in hdmi probe. It works. I have also tested above
> >> >> series for 5420 and 5250 smdk board. There are no issues.
> >> >>
> >> >
> >> > Can you re-check it? I guess platform_driver_register function would
> be
> >> > failed. Actually, I had faced with same issue at hdmi initial work.
> And
> >> then
> >> > only thing we have to discuss  is different buses issue on single hw
> >> device
> >> > if the above case really works fine.
> >> >
> >> Mr. dae,
> >>
> >> I have re-confirmed. It is working fine. No failure during registering
> >> platform
> >> device. Probe hits immediately after registration. I tried 8~9 times.
> >> No failure.
> >> see logs:
> >>
> >> # dmesg | grep -i RSH
> >> [    0.895000] [RSH][hdmi_probe][1719] Starting phy registeration
> >> [    0.900000] [RSH][hdmiphy_platform_device_probe Enter][644]
> >> [    0.905000] [RSH][hdmiphy_platform_device_probe Exit Success][683]
> >> [    0.910000] [RSH][exynos_hdmiphy_driver_register][768] Phy
> >> registeration Success.
> >> [    0.915000] [RSH][hdmi_probe][1729] Phy registeration completed.
> >>
> >
> > Great. I will also re-check it.
> >


Works fine. :) It seems that some patch to avoid duplicate probe was merged
to mainline.
For this, someone had posed it like below but this patch hasn't been merged
to mainline.
 https://lists.ozlabs.org/pipermail/devicetree-discuss/2012-July/017011.html

So it seems that mainline has other solution to avoid that issue. Is there
anyone to give me more detail for this?

> >> > For this, my opinion is to separate the hdmiphy driver into i2c based
> >> and
> >> > platform device based drivers; they include same common header file
> >> > containing phy configuration data. And it makes hdmi driver call
> >> > exynos_hdmiphy_driver_register function in i2c based hdmiphy or
> platform
> >> > device based hdmiphy drivers according to hdmi driver data.
> >> >
> >>
> >> I am fine with it. We can register hdmiphy-i2c or hdmiphy-platform
> >> driver based on the flag from hdmi driver. But we can still keep both
> >> the drivers in exynos_hdmiphy.c. file and let them share most of the
> >> other callbacks.
> >>
> >
> > Is there any mainline driver that keeps two bus drivers; i2c and
> platform
> > device, in one file? I'm not sure that this is a good way. So it seems
> good
> 
> I haven't come across any such mainline driver. I always have this doubt.
> If it is not fitting properly, I will implement 2 different drivers in 2
> files
> (exynos_hdmiphy_i2c.c and exynos_hdmiphy_platform.c).
> 

Let's do it if there is no other opinion. :)

Thanks,
Inki Dae

> Regards,
> Rahul Sharma.
> 
> > to keep two hdmiphy drivers: One is based on i2c bus, and other is based
> on
> > platform device. Anyway, we should keep different type's drivers because
> > Exynos SoC has different type's hdmiphy IP; based on i2c or memory
> mapped
> > IO.
> >
> > Thanks,
> > Inki Dae
> >
> >> regards,
> >> Rahul Sharma.
> >>
> >> > However, we may need to re-design it if the above case is failed.
> >> >
> >> > Thanks,
> >> > Inki Dae
> >> >
> >> >
> >> >> regards,
> >> >> Rahul Sharma.
> >> >>
> >> >> >         return ret;
> >> >> > }
> >> >> >
> >> >> > Thanks,
> >> >> > Inki Dae
> >> >> >
> >> >> >> -----Original Message-----
> >> >> >> From: Rahul Sharma [mailto:rahul.sharma@samsung.com]
> >> >> >> Sent: Friday, August 30, 2013 3:59 PM
> >> >> >> To: linux-samsung-soc@vger.kernel.org; dri-
> >> devel@lists.freedesktop.org
> >> >> >> Cc: kgene.kim@samsung.com; sw0312.kim@samsung.com;
> >> > inki.dae@samsung.com;
> >> >> >> seanpaul@chromium.org; l.stach@pengutronix.de;
> > tomasz.figa@gmail.com;
> >> >> >> s.nawrocki@samsung.com; joshi@samsung.com; r.sh.open@gmail.com;
> >> Rahul
> >> >> >> Sharma
> >> >> >> Subject: [PATCH 0/7] drm/exynos: move hdmiphy related code to
> >> hdmiphy
> >> >> >> driver
> >> >> >>
> >> >> >> Currently, exynos hdmiphy operations and configs are kept
> >> >> >> inside the hdmi driver. Hdmiphy related code is tightly
> >> >> >> coupled with hdmi IP driver.
> >> >> >>
> >> >> >> This series also removes hdmiphy dummy clock for hdmiphy
> >> >> >> and replace it with Phy PMU Control from the hdmiphy driver.
> >> >> >>
> >> >> >> At the end, support for exynos5420 hdmiphy is added to the
> >> >> >> hdmiphy driver which is a platform device.
> >> >> >>
> >> >> >> Drm related paches are based on exynos-drm-next branch at
> >> >> >> git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-
> exynos.git
> >> >> >>
> >> >> >> Arch patches are dependent on
> >> >> >> http://www.mail-archive.com/linux-samsung-
> >> >> >> soc@vger.kernel.org/msg22195.html
> >> >> >>
> >> >> >> Rahul Sharma (7):
> >> >> >>   drm/exynos: move hdmiphy related code to hdmiphy driver
> >> >> >>   drm/exynos: remove dummy hdmiphy clock
> >> >> >>   drm/exynos: add hdmiphy pmu bit control in hdmiphy driver
> >> >> >>   drm/exynos: add support for exynos5420 hdmiphy
> >> >> >>   exynos/drm: fix ddc i2c device probe failure
> >> >> >>   ARM: dts: update hdmiphy dt node for exynos5250
> >> >> >>   ARM: dts: update hdmiphy dt node for exynos5420
> >> >> >>
> >> >> >>  .../devicetree/bindings/video/exynos_hdmi.txt      |    2 +
> >> >> >>  .../devicetree/bindings/video/exynos_hdmiphy.txt   |    6 +
> >> >> >>  arch/arm/boot/dts/exynos5250-smdk5250.dts          |    9 +-
> >> >> >>  arch/arm/boot/dts/exynos5420.dtsi                  |   12 +
> >> >> >>  drivers/gpu/drm/exynos/exynos_ddc.c                |    5 +
> >> >> >>  drivers/gpu/drm/exynos/exynos_drm_hdmi.h           |   13 +
> >> >> >>  drivers/gpu/drm/exynos/exynos_hdmi.c               |  353
> > ++--------
> >> >> >>  drivers/gpu/drm/exynos/exynos_hdmiphy.c            |  738
> >> >> >> +++++++++++++++++++-
> >> >> >>  drivers/gpu/drm/exynos/regs-hdmiphy.h              |   35 +
> >> >> >>  9 files changed, 868 insertions(+), 305 deletions(-)
> >> >> >>  create mode 100644 drivers/gpu/drm/exynos/regs-hdmiphy.h
> >> >> >>
> >> >> >> --
> >> >> >> 1.7.10.4
> >> >> >
> >> >> --
> >> >> To unsubscribe from this list: send the line "unsubscribe linux-
> >> samsung-
> >> >> soc" in
> >> >> the body of a message to majordomo@vger.kernel.org
> >> >> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >> >
> >