Re: [PATCH v4 00/21] qcom: rpmh core and regulator support
On 10/07/2024 13:33, Sumit Garg wrote: Hi Caleb, On Tue, 9 Jul 2024 at 15:04, Caleb Connolly wrote: This series introduces support for the RPMh (Resource Power Manager (hardened)) co-processor and associated regulator driver found on most modern Qualcomm platforms (since ~2017). Even though most regulators are controlled via SPMI, the specific regions on the PMICs for controlling the regulators are restricted and can't be accessed by the Applications Processor (AP/HLOS). Instead, these resources are proxied via the RPMh where they can be voted on by multiple subsystems (Linux, the modem, and the other DSPs). This is done for security (to protect clocks, power domains, and regulators that are specifically relevant for the trustzone) as well as to simplify the management of shared peripherals and to allow for handover of peripherals like the USB controller. For U-Boot, our main concern is the regulators. Since all regulators on the system are controlled via the RPMh, it is necessary to support it to enable USB VBUS on some platforms, and may be needed for other peripherals in the future. Communicating with the RPMh additional requires accessing the cmd-db shared memory region, this contains key/value maps to determine the address of specific resources on the RPMh. Introduce support for the cmd-db, the RPMh framework, and some of the regulators that are necessary to enable USB VBUS on the RB5 development board. These drivers are taken from Linux, then modified and simplified for U-Boot. The original Linux drivers contain heavy optimisations related to multithreading and asynchronous probing, as well as support for idle and suspend states which we don't need to deal with here. This unused code is removed before finally adjusting the drivers to properly build for U-Boot and use its device model. The U-Boot version of the driver supports a single ACTIVE_ONLY TCS and waits for it to be cleared after use. We don't support programming low power states. --- Changes in v4: - Denote original Linux version in bitmap.h patch - Rebased on Linux 6.10-rc6 and re-apply U-Boot changes preserving git history. Allowing for future changes to the Linux drivers to be ported over more easily. - Add missing check to wait for the TCS to be cleared after use (seems we were just racing the RPMh before, oops!). - Fix missing n_modes in pmic5_pldo regulator. - Link to v3: https://lore.kernel.org/r/20240708-b4-qcom-rpmh-v3-0-846cc6c5b...@linaro.org Thanks for doing this, this looks much better from a maintainability perspective. Although I am seeing a lot of patches saying adjustments and build fixes, I think you should have taken care of commits bisect-ability such that individual commits don't cause any build errors. I'll send a v5 which holds off on adding the Kconfig/Makefile parts until after the code is adjusted, this will prevent breaking bisection and still make it easy for us to cherry-pick commits from Linux with some git magic. FWIW, for the series: Acked-by: Sumit Garg Thanks, -Sumit Changes in v3: - Don't call dm_scan_fdt_dev(), since DM core will scan. - Link to v2: https://lore.kernel.org/r/20240708-b4-qcom-rpmh-v2-0-8bc765606...@linaro.org Changes in v2: - Implement Neil's suggestions and fixes for SM8[56]50 - Slightly refactor cmd_db_init() for better abstraction. - Improve logging (printf -> log_err/dev_err) - Add missing error check in rpmh_regulators_bind() - Link to v1: https://lore.kernel.org/r/20240617-b4-qcom-rpmh-v1-0-bd2336923...@linaro.org --- Caleb Connolly (21): linux/bitmap.h: add bitmap_empty helper soc: qcom: import rpmh and cmd-db drivers from Linux soc: qcom: cmd-db: adjust headers for U-Boot soc: qcom: cmd-db: drop unused functions soc: qcom: cmd-db: replace cmd_db_ready() with cmd_db_init() soc: qcom: cmd-db: adjust for U-Boot API soc: qcom: rpmh-rsc: drop unused multi-threading and non-active TCS support soc: qcom: rpmh-rsc: adjust headers for U-Boot soc: qcom: rpmh-rsc: adjust probe for U-Boot soc: qcom: rpmh-rsc: remaining U-Boot API changes soc: qcom: rpmh: adjust headers for U-Boot soc: qcom: rpmh: drop unused functions soc: qcom: rpmh: U-Boot API changes power: regulator: import qcom-rpmh-regulator from Linux power: regulator: adjust headers for U-Boot power: regulator: qcom-rpmh-regulator: port over lineage_range helpers power: regulator: adjust structs for U-Boot power: regulator: qcom-rpmh-regulator: remove unused regulators power: regulator: qcom-rpmh-regulator: port ops to U-Boot power: regulator: qcom-rpmh-regulator: adjust probe for U-Boot qcom_defconfig: enable rpmh regulators configs/qcom_defconfig| 5 + drivers/power/regulator/Kconfig | 8 + drivers/power/regulator/Makefile | 1 + drivers/power/regulator/qcom-rpmh-regulator.c | 544
Re: [PATCH v4 00/21] qcom: rpmh core and regulator support
Hi Caleb, On Tue, 9 Jul 2024 at 15:04, Caleb Connolly wrote: > > This series introduces support for the RPMh (Resource Power Manager > (hardened)) co-processor and associated regulator driver found on most > modern Qualcomm platforms (since ~2017). > > Even though most regulators are controlled via SPMI, the specific > regions on the PMICs for controlling the regulators are restricted and > can't be accessed by the Applications Processor (AP/HLOS). Instead, > these resources are proxied via the RPMh where they can be voted on by > multiple subsystems (Linux, the modem, and the other DSPs). This is done > for security (to protect clocks, power domains, and regulators that are > specifically relevant for the trustzone) as well as to simplify the > management of shared peripherals and to allow for handover of > peripherals like the USB controller. > > For U-Boot, our main concern is the regulators. Since all regulators on > the system are controlled via the RPMh, it is necessary to support it to > enable USB VBUS on some platforms, and may be needed for other > peripherals in the future. > > Communicating with the RPMh additional requires accessing the cmd-db > shared memory region, this contains key/value maps to determine the > address of specific resources on the RPMh. > > Introduce support for the cmd-db, the RPMh framework, and some of the > regulators that are necessary to enable USB VBUS on the RB5 development > board. > > These drivers are taken from Linux, then modified and simplified for > U-Boot. The original Linux drivers contain heavy optimisations related > to multithreading and asynchronous probing, as well as support for idle > and suspend states which we don't need to deal with here. This unused > code is removed before finally adjusting the drivers to properly build > for U-Boot and use its device model. > > The U-Boot version of the driver supports a single ACTIVE_ONLY TCS and > waits for it to be cleared after use. We don't support programming > low power states. > > --- > Changes in v4: > - Denote original Linux version in bitmap.h patch > - Rebased on Linux 6.10-rc6 and re-apply U-Boot changes preserving git > history. Allowing for future changes to the Linux drivers to be ported > over more easily. > - Add missing check to wait for the TCS to be cleared after use (seems > we were just racing the RPMh before, oops!). > - Fix missing n_modes in pmic5_pldo regulator. > - Link to v3: > https://lore.kernel.org/r/20240708-b4-qcom-rpmh-v3-0-846cc6c5b...@linaro.org Thanks for doing this, this looks much better from a maintainability perspective. Although I am seeing a lot of patches saying adjustments and build fixes, I think you should have taken care of commits bisect-ability such that individual commits don't cause any build errors. FWIW, for the series: Acked-by: Sumit Garg -Sumit > > Changes in v3: > - Don't call dm_scan_fdt_dev(), since DM core will scan. > - Link to v2: > https://lore.kernel.org/r/20240708-b4-qcom-rpmh-v2-0-8bc765606...@linaro.org > > Changes in v2: > - Implement Neil's suggestions and fixes for SM8[56]50 > - Slightly refactor cmd_db_init() for better abstraction. > - Improve logging (printf -> log_err/dev_err) > - Add missing error check in rpmh_regulators_bind() > - Link to v1: > https://lore.kernel.org/r/20240617-b4-qcom-rpmh-v1-0-bd2336923...@linaro.org > > --- > Caleb Connolly (21): > linux/bitmap.h: add bitmap_empty helper > soc: qcom: import rpmh and cmd-db drivers from Linux > soc: qcom: cmd-db: adjust headers for U-Boot > soc: qcom: cmd-db: drop unused functions > soc: qcom: cmd-db: replace cmd_db_ready() with cmd_db_init() > soc: qcom: cmd-db: adjust for U-Boot API > soc: qcom: rpmh-rsc: drop unused multi-threading and non-active TCS > support > soc: qcom: rpmh-rsc: adjust headers for U-Boot > soc: qcom: rpmh-rsc: adjust probe for U-Boot > soc: qcom: rpmh-rsc: remaining U-Boot API changes > soc: qcom: rpmh: adjust headers for U-Boot > soc: qcom: rpmh: drop unused functions > soc: qcom: rpmh: U-Boot API changes > power: regulator: import qcom-rpmh-regulator from Linux > power: regulator: adjust headers for U-Boot > power: regulator: qcom-rpmh-regulator: port over lineage_range helpers > power: regulator: adjust structs for U-Boot > power: regulator: qcom-rpmh-regulator: remove unused regulators > power: regulator: qcom-rpmh-regulator: port ops to U-Boot > power: regulator: qcom-rpmh-regulator: adjust probe for U-Boot > qcom_defconfig: enable rpmh regulators > > configs/qcom_defconfig| 5 + > drivers/power/regulator/Kconfig | 8 + > drivers/power/regulator/Makefile | 1 + > drivers/power/regulator/qcom-rpmh-regulator.c | 544 > ++ > drivers/soc/Kconfig | 1 + > drivers/soc/Makefile | 1
