| Message ID | 20220714124333.27643-1-johan+linaro@kernel.org (mailing list archive) |
|---|---|
| Headers | show |
| Series | phy: qcom,qmp: fix dt-bindings and deprecate lane suffix | expand |
On Thu, Jul 14, 2022 at 02:43:03PM +0200, Johan Hovold wrote: > When adding support for SC8280XP to the QMP PHY driver I noticed that > the PHY provider child node was not described by the current DT schema. > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > the clock name just to match the current "pipe0" name so I decided to > deprecate the unnecessary suffix in the current binding instead. > > To be able to add the missing child-node schema and handle device > specifics like additional PIPE clocks, it quickly became obvious that > the binding needs to be split up. > > This series clean up and fixes some issue with the current schema before > splitting it up in separate schemas for PCIe, UFS and USB and adding > missing parts like the child PHY provider nodes. > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > PHY that actually provides more than one PHY per IP block. Note that the > "lane" suffix is still unnecessary and misleading. > > The final patches add support for the updated binding to the (recently > split up) PHY drivers. Included is also a related combo PHY cleanup. Hi Vinod, any chance of getting these into 5.20? Note that there'll be an -rc8 on Sunday. Johan
Hi Vinod, On Mon, Jul 18, 2022 at 02:17:36PM +0200, Johan Hovold wrote: > On Thu, Jul 14, 2022 at 02:43:03PM +0200, Johan Hovold wrote: > > When adding support for SC8280XP to the QMP PHY driver I noticed that > > the PHY provider child node was not described by the current DT schema. > > > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > > the clock name just to match the current "pipe0" name so I decided to > > deprecate the unnecessary suffix in the current binding instead. > > > > To be able to add the missing child-node schema and handle device > > specifics like additional PIPE clocks, it quickly became obvious that > > the binding needs to be split up. > > > > This series clean up and fixes some issue with the current schema before > > splitting it up in separate schemas for PCIe, UFS and USB and adding > > missing parts like the child PHY provider nodes. > > > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > > PHY that actually provides more than one PHY per IP block. Note that the > > "lane" suffix is still unnecessary and misleading. > > > > The final patches add support for the updated binding to the (recently > > split up) PHY drivers. Included is also a related combo PHY cleanup. > > Hi Vinod, > > any chance of getting these into 5.20? > > Note that there'll be an -rc8 on Sunday. Have you had a chance to look at this series yet? Johan
On 18-07-22, 14:17, Johan Hovold wrote: > On Thu, Jul 14, 2022 at 02:43:03PM +0200, Johan Hovold wrote: > > When adding support for SC8280XP to the QMP PHY driver I noticed that > > the PHY provider child node was not described by the current DT schema. > > > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > > the clock name just to match the current "pipe0" name so I decided to > > deprecate the unnecessary suffix in the current binding instead. > > > > To be able to add the missing child-node schema and handle device > > specifics like additional PIPE clocks, it quickly became obvious that > > the binding needs to be split up. > > > > This series clean up and fixes some issue with the current schema before > > splitting it up in separate schemas for PCIe, UFS and USB and adding > > missing parts like the child PHY provider nodes. > > > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > > PHY that actually provides more than one PHY per IP block. Note that the > > "lane" suffix is still unnecessary and misleading. > > > > The final patches add support for the updated binding to the (recently > > split up) PHY drivers. Included is also a related combo PHY cleanup. > > Hi Vinod, > > any chance of getting these into 5.20? > > Note that there'll be an -rc8 on Sunday. Sorry phy tree goes thru char-misc and it was already sent and closed, so couldnt do the needful
On 14-07-22, 14:43, Johan Hovold wrote: > When adding support for SC8280XP to the QMP PHY driver I noticed that > the PHY provider child node was not described by the current DT schema. > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > the clock name just to match the current "pipe0" name so I decided to > deprecate the unnecessary suffix in the current binding instead. > > To be able to add the missing child-node schema and handle device > specifics like additional PIPE clocks, it quickly became obvious that > the binding needs to be split up. > > This series clean up and fixes some issue with the current schema before > splitting it up in separate schemas for PCIe, UFS and USB and adding > missing parts like the child PHY provider nodes. > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > PHY that actually provides more than one PHY per IP block. Note that the > "lane" suffix is still unnecessary and misleading. > > The final patches add support for the updated binding to the (recently > split up) PHY drivers. Included is also a related combo PHY cleanup. This fails at patch 2 for me on v6.0-rc1, please rebase and resend Thanks
On Tue, Aug 30, 2022 at 10:36:43AM +0530, Vinod Koul wrote: > On 14-07-22, 14:43, Johan Hovold wrote: > > When adding support for SC8280XP to the QMP PHY driver I noticed that > > the PHY provider child node was not described by the current DT schema. > > > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > > the clock name just to match the current "pipe0" name so I decided to > > deprecate the unnecessary suffix in the current binding instead. > > > > To be able to add the missing child-node schema and handle device > > specifics like additional PIPE clocks, it quickly became obvious that > > the binding needs to be split up. > > > > This series clean up and fixes some issue with the current schema before > > splitting it up in separate schemas for PCIe, UFS and USB and adding > > missing parts like the child PHY provider nodes. > > > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > > PHY that actually provides more than one PHY per IP block. Note that the > > "lane" suffix is still unnecessary and misleading. > > > > The final patches add support for the updated binding to the (recently > > split up) PHY drivers. Included is also a related combo PHY cleanup. > > This fails at patch 2 for me on v6.0-rc1, please rebase and resend Are you sure you haven't applied anything to your local tree that causes this? I just tried fetching the v3 series from lore and it applies just fine on top of 6.0-rc1. Note that if you've added a new compatible string locally, the second patch which sorts the compatible strings is likely to fail to apply. Johan
On Tue, Aug 30, 2022 at 10:35:24AM +0530, Vinod Koul wrote: > On 18-07-22, 14:17, Johan Hovold wrote: > > On Thu, Jul 14, 2022 at 02:43:03PM +0200, Johan Hovold wrote: > > > When adding support for SC8280XP to the QMP PHY driver I noticed that > > > the PHY provider child node was not described by the current DT schema. > > > > > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > > > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > > > the clock name just to match the current "pipe0" name so I decided to > > > deprecate the unnecessary suffix in the current binding instead. > > > > > > To be able to add the missing child-node schema and handle device > > > specifics like additional PIPE clocks, it quickly became obvious that > > > the binding needs to be split up. > > > > > > This series clean up and fixes some issue with the current schema before > > > splitting it up in separate schemas for PCIe, UFS and USB and adding > > > missing parts like the child PHY provider nodes. > > > > > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > > > PHY that actually provides more than one PHY per IP block. Note that the > > > "lane" suffix is still unnecessary and misleading. > > > > > > The final patches add support for the updated binding to the (recently > > > split up) PHY drivers. Included is also a related combo PHY cleanup. > > > > Hi Vinod, > > > > any chance of getting these into 5.20? > > > > Note that there'll be an -rc8 on Sunday. > > Sorry phy tree goes thru char-misc and it was already sent and closed, > so couldnt do the needful Yeah, it was a bit of a wild shot, but I noticed you hadn't actually sent your PR to Greg yet when I pinged you. Johan
On 30-08-22, 09:01, Johan Hovold wrote: > On Tue, Aug 30, 2022 at 10:36:43AM +0530, Vinod Koul wrote: > > On 14-07-22, 14:43, Johan Hovold wrote: > > > When adding support for SC8280XP to the QMP PHY driver I noticed that > > > the PHY provider child node was not described by the current DT schema. > > > > > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > > > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > > > the clock name just to match the current "pipe0" name so I decided to > > > deprecate the unnecessary suffix in the current binding instead. > > > > > > To be able to add the missing child-node schema and handle device > > > specifics like additional PIPE clocks, it quickly became obvious that > > > the binding needs to be split up. > > > > > > This series clean up and fixes some issue with the current schema before > > > splitting it up in separate schemas for PCIe, UFS and USB and adding > > > missing parts like the child PHY provider nodes. > > > > > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > > > PHY that actually provides more than one PHY per IP block. Note that the > > > "lane" suffix is still unnecessary and misleading. > > > > > > The final patches add support for the updated binding to the (recently > > > split up) PHY drivers. Included is also a related combo PHY cleanup. > > > > This fails at patch 2 for me on v6.0-rc1, please rebase and resend > > Are you sure you haven't applied anything to your local tree that causes > this? Pretty sure :-) > > I just tried fetching the v3 series from lore and it applies just fine > on top of 6.0-rc1. > > Note that if you've added a new compatible string locally, the second > patch which sorts the compatible strings is likely to fail to apply. At that time no, now I think I have patch or so ... Tree should be pushed in a bit, you can check
On Tue, Aug 30, 2022 at 12:58:21PM +0530, Vinod Koul wrote: > On 30-08-22, 09:01, Johan Hovold wrote: > > On Tue, Aug 30, 2022 at 10:36:43AM +0530, Vinod Koul wrote: > > > On 14-07-22, 14:43, Johan Hovold wrote: > > > > When adding support for SC8280XP to the QMP PHY driver I noticed that > > > > the PHY provider child node was not described by the current DT schema. > > > > > > > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > > > > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > > > > the clock name just to match the current "pipe0" name so I decided to > > > > deprecate the unnecessary suffix in the current binding instead. > > > > > > > > To be able to add the missing child-node schema and handle device > > > > specifics like additional PIPE clocks, it quickly became obvious that > > > > the binding needs to be split up. > > > > > > > > This series clean up and fixes some issue with the current schema before > > > > splitting it up in separate schemas for PCIe, UFS and USB and adding > > > > missing parts like the child PHY provider nodes. > > > > > > > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > > > > PHY that actually provides more than one PHY per IP block. Note that the > > > > "lane" suffix is still unnecessary and misleading. > > > > > > > > The final patches add support for the updated binding to the (recently > > > > split up) PHY drivers. Included is also a related combo PHY cleanup. > > > > > > This fails at patch 2 for me on v6.0-rc1, please rebase and resend > > > > Are you sure you haven't applied anything to your local tree that causes > > this? > > Pretty sure :-) Hmm. But nothing had changed in 6.0-rc1 and it still applies on a clean 6.0-rc1 as expected here. Would you mind trying again? git checkout -b tmp-branch v6.0-rc1 b4 am 20220714124333.27643-1-johan+linaro@kernel.org git am ./v3_20220714_johan_linaro_phy_qcom_qmp_fix_dt_bindings_and_deprecate_lane_suffix.mbx > > I just tried fetching the v3 series from lore and it applies just fine > > on top of 6.0-rc1. > > > > Note that if you've added a new compatible string locally, the second > > patch which sorts the compatible strings is likely to fail to apply. > > At that time no, now I think I have patch or so ... Tree should be > pushed in a bit, you can check Which tree would that be? The linux-phy tree next branch is still at -rc1: https://git.kernel.org/pub/scm/linux/kernel/git/phy/linux-phy.git/ Johan
On 30-08-22, 09:43, Johan Hovold wrote: > On Tue, Aug 30, 2022 at 12:58:21PM +0530, Vinod Koul wrote: > > On 30-08-22, 09:01, Johan Hovold wrote: > > > On Tue, Aug 30, 2022 at 10:36:43AM +0530, Vinod Koul wrote: > > > > On 14-07-22, 14:43, Johan Hovold wrote: > > > > > When adding support for SC8280XP to the QMP PHY driver I noticed that > > > > > the PHY provider child node was not described by the current DT schema. > > > > > > > > > > The SC8280XP PHYs also need a second fixed-divider PIPE clock > > > > > ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to > > > > > the clock name just to match the current "pipe0" name so I decided to > > > > > deprecate the unnecessary suffix in the current binding instead. > > > > > > > > > > To be able to add the missing child-node schema and handle device > > > > > specifics like additional PIPE clocks, it quickly became obvious that > > > > > the binding needs to be split up. > > > > > > > > > > This series clean up and fixes some issue with the current schema before > > > > > splitting it up in separate schemas for PCIe, UFS and USB and adding > > > > > missing parts like the child PHY provider nodes. > > > > > > > > > > The MSM8996 PCIe PHY gets its own schema as this is the only non-combo > > > > > PHY that actually provides more than one PHY per IP block. Note that the > > > > > "lane" suffix is still unnecessary and misleading. > > > > > > > > > > The final patches add support for the updated binding to the (recently > > > > > split up) PHY drivers. Included is also a related combo PHY cleanup. > > > > > > > > This fails at patch 2 for me on v6.0-rc1, please rebase and resend > > > > > > Are you sure you haven't applied anything to your local tree that causes > > > this? > > > > Pretty sure :-) > > Hmm. But nothing had changed in 6.0-rc1 and it still applies on a clean > 6.0-rc1 as expected here. > > Would you mind trying again? > > git checkout -b tmp-branch v6.0-rc1 > b4 am 20220714124333.27643-1-johan+linaro@kernel.org > git am ./v3_20220714_johan_linaro_phy_qcom_qmp_fix_dt_bindings_and_deprecate_lane_suffix.mbx > > > > I just tried fetching the v3 series from lore and it applies just fine > > > on top of 6.0-rc1. > > > > > > Note that if you've added a new compatible string locally, the second > > > patch which sorts the compatible strings is likely to fail to apply. > > > > At that time no, now I think I have patch or so ... Tree should be > > pushed in a bit, you can check > > Which tree would that be? The linux-phy tree next branch is still at -rc1: > > https://git.kernel.org/pub/scm/linux/kernel/git/phy/linux-phy.git/ It is updated now, my tests were running
On 30/08/2022 10:01, Johan Hovold wrote: > On Tue, Aug 30, 2022 at 10:36:43AM +0530, Vinod Koul wrote: >> On 14-07-22, 14:43, Johan Hovold wrote: [skipped] >>> The final patches add support for the updated binding to the (recently >>> split up) PHY drivers. Included is also a related combo PHY cleanup. >> >> This fails at patch 2 for me on v6.0-rc1, please rebase and resend > > Are you sure you haven't applied anything to your local tree that causes > this? When doing the QMP split, we ended up in a similar situation if I remember correctly. Most probably the easiest way is to provide a branch for Vinod to pull in addition to the patches being sent to ML. > I just tried fetching the v3 series from lore and it applies just fine > on top of 6.0-rc1. > > Note that if you've added a new compatible string locally, the second > patch which sorts the compatible strings is likely to fail to apply. > > Johan
On Tue, Aug 30, 2022 at 12:32:04PM +0300, Dmitry Baryshkov wrote: > On 30/08/2022 10:01, Johan Hovold wrote: > > On Tue, Aug 30, 2022 at 10:36:43AM +0530, Vinod Koul wrote: > >> On 14-07-22, 14:43, Johan Hovold wrote: > > [skipped] > > >>> The final patches add support for the updated binding to the (recently > >>> split up) PHY drivers. Included is also a related combo PHY cleanup. > >> > >> This fails at patch 2 for me on v6.0-rc1, please rebase and resend > > > > Are you sure you haven't applied anything to your local tree that causes > > this? > > When doing the QMP split, we ended up in a similar situation if I > remember correctly. Most probably the easiest way is to provide a branch > for Vinod to pull in addition to the patches being sent to ML. Hopefully v4 will just work now. > > I just tried fetching the v3 series from lore and it applies just fine > > on top of 6.0-rc1. > > > > Note that if you've added a new compatible string locally, the second > > patch which sorts the compatible strings is likely to fail to apply. Johan
When adding support for SC8280XP to the QMP PHY driver I noticed that the PHY provider child node was not described by the current DT schema. The SC8280XP PHYs also need a second fixed-divider PIPE clock ("pipediv2") and I didn't want to have to add a bogus "lane" suffix to the clock name just to match the current "pipe0" name so I decided to deprecate the unnecessary suffix in the current binding instead. To be able to add the missing child-node schema and handle device specifics like additional PIPE clocks, it quickly became obvious that the binding needs to be split up. This series clean up and fixes some issue with the current schema before splitting it up in separate schemas for PCIe, UFS and USB and adding missing parts like the child PHY provider nodes. The MSM8996 PCIe PHY gets its own schema as this is the only non-combo PHY that actually provides more than one PHY per IP block. Note that the "lane" suffix is still unnecessary and misleading. The final patches add support for the updated binding to the (recently split up) PHY drivers. Included is also a related combo PHY cleanup. Johan Changes in v3 - rebase on linux-next which has a new binding for IPQ8074 - fix git-bisect breakage due to removal of an unused variable one patch too soon (Krzysztof) - replace one Fixes tag with reference in commit message (Krzysztof) - drop two redundant minItems (Krzysztof) - fix two Fixes tags that lacked the actual tag - add more ack and review tags from Krzysztof Changes in v2 - squash split + cleanup + example patches (Krzysztof) - deprecate clock-names instead of dropping suffix (Krzysztof) - deprecate reset-names instead of dropping suffix (Krzysztof) - flatten child reg if/then schemas (Krzysztof) - add back optional vddp-ref-clk to all bindings even though it likely only applies to MSM8996/98 UFS (Krzysztof) - add missing sc7180 schema to USB binding - misc clean ups - shorten or drop descriptions - drop quotes around $id and $schema (Krzysztof) - use maxItems with clock-output-names - combine two USB clock+reset schemas - add Reviewed-by/Acked-by tags Johan Hovold (30): dt-bindings: phy: qcom,qmp: fix bogus clock-cells property dt-bindings: phy: qcom,qmp: sort compatible strings dt-bindings: phy: qcom,qmp: drop redundant descriptions dt-bindings: phy: qcom,qmp: fix child node description dt-bindings: phy: qcom,qmp: clean up descriptions dt-bindings: phy: qcom,qmp: clean up example dt-bindings: phy: qcom,qmp: drop child-node comment dt-bindings: phy: add qcom,msm8996-qmp-pcie-phy schema dt-bindings: phy: qcom,msm8996-qmp-pcie: add missing child node schema dt-bindings: phy: qcom,msm8996-qmp-pcie: deprecate PIPE clock names dt-bindings: phy: qcom,msm8996-qmp-pcie: deprecate reset names dt-bindings: phy: add QMP PCIe PHY schema dt-bindings: phy: qcom,qmp-pcie: add missing child node schema dt-bindings: phy: qcom,qmp-pcie: deprecate PIPE clock name dt-bindings: phy: add QMP UFS PHY schema dt-bindings: phy: qcom,qmp-ufs: add missing SM8450 clock dt-bindings: phy: qcom,qmp-ufs: add missing SM8150 power domain dt-bindings: phy: qcom,qmp-ufs: add missing child node schema dt-bindings: phy: add QMP USB PHY schema dt-bindings: phy: qcom,qmp-usb: add missing child node schema dt-bindings: phy: qcom,qmp-usb: deprecate PIPE clock name dt-bindings: phy: qcom,qmp-usb: add missing qcom,sc7180-qmp-usb3-phy schema dt-bindings: phy: qcom,qmp-usb3-dp: fix bogus clock-cells property dt-bindings: phy: qcom,qmp-usb3-dp: deprecate USB PIPE clock name phy: qcom-qmp-pcie: drop pipe clock lane suffix phy: qcom-qmp-combo: drop unused lane reset phy: qcom-qmp-combo: drop pipe clock lane suffix phy: qcom-qmp-pcie-msm8996: drop pipe clock lane suffix phy: qcom-qmp-pcie-msm8996: drop reset lane suffix phy: qcom-qmp-usb: drop pipe clock lane suffix .../phy/qcom,msm8996-qmp-pcie-phy.yaml | 189 +++++++ .../bindings/phy/qcom,qmp-pcie-phy.yaml | 296 +++++++++++ .../devicetree/bindings/phy/qcom,qmp-phy.yaml | 502 ------------------ .../bindings/phy/qcom,qmp-ufs-phy.yaml | 239 +++++++++ .../bindings/phy/qcom,qmp-usb-phy.yaml | 385 ++++++++++++++ .../bindings/phy/qcom,qmp-usb3-dp-phy.yaml | 8 +- drivers/phy/qualcomm/phy-qcom-qmp-combo.c | 6 +- .../phy/qualcomm/phy-qcom-qmp-pcie-msm8996.c | 8 +- drivers/phy/qualcomm/phy-qcom-qmp-pcie.c | 4 +- drivers/phy/qualcomm/phy-qcom-qmp-usb.c | 4 +- 10 files changed, 1115 insertions(+), 526 deletions(-) create mode 100644 Documentation/devicetree/bindings/phy/qcom,msm8996-qmp-pcie-phy.yaml create mode 100644 Documentation/devicetree/bindings/phy/qcom,qmp-pcie-phy.yaml delete mode 100644 Documentation/devicetree/bindings/phy/qcom,qmp-phy.yaml create mode 100644 Documentation/devicetree/bindings/phy/qcom,qmp-ufs-phy.yaml create mode 100644 Documentation/devicetree/bindings/phy/qcom,qmp-usb-phy.yaml