| Message ID | 20200904191830.387296-1-jagan@amarulasolutions.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | extcon: Add Type-C Virtual PD | expand |
Hi Jagan, Am Freitag, 4. September 2020, 21:18:27 CEST schrieb Jagan Teki: > USB Type-C protocol supports various modes of operations > includes PD, USB3, and Altmode. If the platform design > supports a Type-C connector then configuring these modes > can be done via enumeration. > > However, there are some platforms that design these modes > of operations as separate protocol connectors like design > Display Port from on-chip USB3 controller. So accessing > Type-C Altmode Display Port via onboard Display Port > connector instead of a Type-C connector. > > These kinds of platforms require an explicit extcon driver > in order to handle Power Delivery and Port Detection. > > This series support this Type-C Virtual PD and enable the > same in ROCK Pi 4C SBC. > > Any inputs? I tend to disagree on the design via an extcon. That the Rockchip rk3399 currently carries that extcon thingy is unfortunate and only works for ChromeOS devices based on the rk3399. The kernel now has a real type-c framework so we should not extend this extcon hack any further, because that will make it even harder to roll back later. Also simply because other Rockchip boards currently can't really make use of type-c due to this, as they use the fsusb302 phys directly connected. ChromeOS actually spend some time to make the cros-ec pd part of the type-c framework if I remember correctly, so a viable battle plan would be to: (1) move the Rockchip type-c phy driver to actually be part of the type-c framework, with the extcon being a deprecated fallback for old DTs. (2) implement your gpio-altmode as part of the type-c framework (which may even already exist) In short, please don't extend the rk3399 type-c extcon hack. Thanks Heiko
Hi Heiko, On Sat, Sep 5, 2020 at 2:53 AM Heiko Stübner <heiko@sntech.de> wrote: > > Hi Jagan, > > Am Freitag, 4. September 2020, 21:18:27 CEST schrieb Jagan Teki: > > USB Type-C protocol supports various modes of operations > > includes PD, USB3, and Altmode. If the platform design > > supports a Type-C connector then configuring these modes > > can be done via enumeration. > > > > However, there are some platforms that design these modes > > of operations as separate protocol connectors like design > > Display Port from on-chip USB3 controller. So accessing > > Type-C Altmode Display Port via onboard Display Port > > connector instead of a Type-C connector. > > > > These kinds of platforms require an explicit extcon driver > > in order to handle Power Delivery and Port Detection. > > > > This series support this Type-C Virtual PD and enable the > > same in ROCK Pi 4C SBC. > > > > Any inputs? > > I tend to disagree on the design via an extcon. Okay, are you disagree with the extcon extension for fusb chips routing or entire rk3399 designs? I totally agree with this point of bypassing extcon for the designs which has fusb302 chips. My only concern is about designs that don't have fusb chips - for example rock-pi-4. Designs that do have fusb302 chips, has dynamic possibilities to identify data roles, like detecting Altmode DP via Type-C connector whereas designs that don't have fusb302 or any type-c chip need static identification of data role, polarity, and ss for detecting direct DP port ie where virtual-pd is useful. Look like we have two potential issues of handling DP on rk3399 here, let me know if you think these non-type-c chips designs also possible to detect w/o extcon? Jagan.
On Fri, Sep 04, 2020 at 11:23:21PM +0200, Heiko Stübner wrote: > Hi Jagan, > > Am Freitag, 4. September 2020, 21:18:27 CEST schrieb Jagan Teki: > > USB Type-C protocol supports various modes of operations > > includes PD, USB3, and Altmode. If the platform design > > supports a Type-C connector then configuring these modes > > can be done via enumeration. > > > > However, there are some platforms that design these modes > > of operations as separate protocol connectors like design > > Display Port from on-chip USB3 controller. So accessing > > Type-C Altmode Display Port via onboard Display Port > > connector instead of a Type-C connector. > > > > These kinds of platforms require an explicit extcon driver > > in order to handle Power Delivery and Port Detection. > > > > This series support this Type-C Virtual PD and enable the > > same in ROCK Pi 4C SBC. > > > > Any inputs? > > I tend to disagree on the design via an extcon. I don't accept new extcon bindings or users of it either. It's a poorly thought out collection of Linux driver properties. Use the usb connector binding. > > That the Rockchip rk3399 currently carries that extcon thingy is unfortunate > and only works for ChromeOS devices based on the rk3399. > > The kernel now has a real type-c framework so we should not extend this > extcon hack any further, because that will make it even harder to roll back > later. Also simply because other Rockchip boards currently can't really make > use of type-c due to this, as they use the fsusb302 phys directly connected. > > ChromeOS actually spend some time to make the cros-ec pd part of the type-c > framework if I remember correctly, so a viable battle plan would be to: > > (1) move the Rockchip type-c phy driver to actually be part of the type-c > framework, with the extcon being a deprecated fallback for old DTs. > (2) implement your gpio-altmode as part of the type-c framework > (which may even already exist) > > > In short, please don't extend the rk3399 type-c extcon hack. > > Thanks > Heiko > > > >
Hi Rob, On 9/15/20 8:15 AM, Rob Herring wrote: > On Fri, Sep 04, 2020 at 11:23:21PM +0200, Heiko Stübner wrote: >> Hi Jagan, >> >> Am Freitag, 4. September 2020, 21:18:27 CEST schrieb Jagan Teki: >>> USB Type-C protocol supports various modes of operations >>> includes PD, USB3, and Altmode. If the platform design >>> supports a Type-C connector then configuring these modes >>> can be done via enumeration. >>> >>> However, there are some platforms that design these modes >>> of operations as separate protocol connectors like design >>> Display Port from on-chip USB3 controller. So accessing >>> Type-C Altmode Display Port via onboard Display Port >>> connector instead of a Type-C connector. >>> >>> These kinds of platforms require an explicit extcon driver >>> in order to handle Power Delivery and Port Detection. >>> >>> This series support this Type-C Virtual PD and enable the >>> same in ROCK Pi 4C SBC. >>> >>> Any inputs? >> >> I tend to disagree on the design via an extcon. > > I don't accept new extcon bindings or users of it either. It's a > poorly thought out collection of Linux driver properties. Use the usb > connector binding. I agree the prior binding format of extcon with 'extcon' property name is not proper. So, I'll try to bind new style. But, I have a question. Do you have the objection of the use of extcon like 'users of it either' of your comment? > >> >> That the Rockchip rk3399 currently carries that extcon thingy is unfortunate >> and only works for ChromeOS devices based on the rk3399. >> >> The kernel now has a real type-c framework so we should not extend this >> extcon hack any further, because that will make it even harder to roll back >> later. Also simply because other Rockchip boards currently can't really make >> use of type-c due to this, as they use the fsusb302 phys directly connected. >> >> ChromeOS actually spend some time to make the cros-ec pd part of the type-c >> framework if I remember correctly, so a viable battle plan would be to: >> >> (1) move the Rockchip type-c phy driver to actually be part of the type-c >> framework, with the extcon being a deprecated fallback for old DTs. >> (2) implement your gpio-altmode as part of the type-c framework >> (which may even already exist) >> >> >> In short, please don't extend the rk3399 type-c extcon hack. >> >> Thanks >> Heiko >> >> >> >> > >