On Wed, Jul 19, 2023 at 2:05 PM Ying Liu <victor.liu@xxxxxxx> wrote: > > On Tuesday, July 18, 2023 6:51 PM Jagan Teki <jagan@xxxxxxxxxxxxxxxxxxxx> wrote: > > > > Hi, > > Hi, > > > > > On Tue, Jul 18, 2023 at 3:19 PM Ying Liu <victor.liu@xxxxxxx> wrote: > > > > > > On Tuesday, July 18, 2023 5:35 PM Jagan Teki > > <jagan@xxxxxxxxxxxxxxxxxxxx> wrote: > > > > > > > > > > > > > > Hi Jagan, > > > > > > > > > > On Monday, July 17, 2023 2:44 PM Jagan Teki > > > > <jagan@xxxxxxxxxxxxxxxxxxxx> wrote: > > > > > > On Mon, Jul 17, 2023 at 11:44 AM Liu Ying <victor.liu@xxxxxxx> > > wrote: > > > > > > > > > > > > > > Freescale i.MX93 SoC embeds a Synopsys Designware MIPI DSI host > > > > > > > controller and a Synopsys Designware MIPI DPHY. Some > > configurations > > > > > > > and extensions to them are controlled by i.MX93 media blk-ctrl. > > > > > > > > > > > > > > Add a DRM bridge for i.MX93 MIPI DSI by using existing DW MIPI DSI > > > > > > > bridge helpers and implementing i.MX93 MIPI DSI specific > > extensions. > > > > > > > > > > > > I think the better way would add compatibility to be part of existing > > > > > > dw-mipi-dsi.c with specific driver data. This way it avoids all the > > > > > > platform-related helpers(extensions) and makes the driver generic to > > > > > > all SoCs which use DW DSI IP. It would be a straightforward change as > > > > > > the imx93 drm pipeline already supports bridge topology. > > > > > > > > > > The platform-related stuff is handed over to dw-mipi-dsi.c via struct > > > > > dw_mipi_dsi_plat_data as an argument of dw_mipi_dsi_probe(). It > > looks > > > > > ok for vendor drivers to call dw_mipi_dsi_probe() to set the platform- > > > > related > > > > > information(rockchip, meson and stm do that), like pdata.phy_ops and > > > > > pdata.host_ops. > > > > > > > > I understand this topology of having soc-platform drivers with > > > > dw-mipi-dsi bridge. What I'm suggesting is to not add a soc-platform > > > > driver for imx93 instead add add support directly on dw-mipi-dsi > > > > bridge. > > > > > > It seems that directly supporting i.MX93 MIPI DSI in dw-mipi-dsi.c is > > > not feasible. The main reason is that struct dw_mipi_dsi_plat_data > > > contains phy_ops and each vendor driver provides very different > > > methods for phy, while... > > > > Cannot this phy_ops goes to PHY core somewhere around and even it is > > possible to add via driver data for imx93 by untouching existing > > handling? I know it is not as direct as I describe but it is worth > > maintaining the driver this way to keep control of the future > > soc-drivers to include in dw-mipi-dsi instead of handling platform > > code separately. > > Like I said, each vendor driver provides very different methods for phy. > The reason is that phy IPs are different and SoC integration things are > handled via struct dw_mipi_dsi_phy_ops. Meson calls devm_phy_get() > to get a phy. Rockchip calls devm_phy_create() to create a phy. Meson, > rockchip and stm have their own struct dw_mipi_dsi_phy_ops > implementations, same to i.MX93. Different pixel format control is > implemented in meson's and i.MX93's ->init() callback. Meson additionally > handles clocks in ->init() callback. > > Generally speaking, it's a no-go to put phy_ops into PHY core for all the > vendors, IMHO. > > In particular, i.MX93 MIPI DPHY's PLL can be fully controlled by media > blk-ctrl(as a syscon) thru the PLL's SoC control interface, while PHY > registers can be controlled by DW MIPI DSI testdin/out control interface > alternatively including a part of the PLL registers. So, adding a PHY > driver for i.MX93 MIPI DPHY doesn't make sense since the PLL controlled > by media blk-ctrl doesn't constitute a PHY. Instead, dw_mipi_dsi_phy_ops > may cover all the PHY control well. > > From my PoV, w/wo this series, dw-mipi-dsi.c looks fine to keep being > generic and easy to maintain. All vendor drivers do is to handle platform > specific stuff, which is good. Thanks for the explanation. I agreed with on you most of the points from the perspective of PoV which are difficult to FIT into a common bridge. However, I'm still believing that the having bridge does only the bridge job, and keeping the PHY or other PoV-specific stuff on respective driver subsystems would be great synergy for the bridge drivers or any common IP driver's point-of-view. Anyway, I might not control i.MX93 PoV stuff, but I can try on it for the new DSI which uses this IP, and keep you in CC if it is a possible land in the near future. > > > > > > > > > > > > > > > > > > > > dw-mipi-dsi.c is generic w/wo this patch series. > > > > > > > > > > Can you elaborate more about adding compatibility to be part of > > existing > > > > > dw-mipi-dsi.c with specific driver data? I don't see clear approach to do > > > > > that. > > > > > > > > Please check the above comments, an example of samsung-dsim.c > > > > > > ... it seems that samsung-dsim.c uses struct samsung_dsim_driver_data to > > > differential platform information and it doesn't contain any callback, which > > > means comparing to DW MIPI DSI, Samsung DSIM shows more consistency > > > across vendor SoCs from HW IP/SoC integration PoV. > > > > Yes, but is it possible to adjust struct according to DW MIPI DSI. > > Looking at the vendor drivers, they show a lot diversity, which cannot be > parameterized into a struct like the samsung dsim driver does. > > struct dw_mipi_dsi_plat_data hides all platform specific information from > dw-mipi-dsi.c well, IMHO. Okay. Thanks, Jagan.
