Voltage and current regulator subsystem (PMIC framework)
A core kernel subsystem that manages the voltage and current rails supplied by power-management ICs (PMICs) and discrete regulators on essentially every modern SoC-based system, from phones to embedded i.MX and MediaTek boards. Individual chip drivers (MT6360, PCA9450, PF0900, TPS68470, and many more) plug in here so other drivers can request and scale their power supplies.
recommendation
It should stay because this is the kernel's central framework for managing power-management ICs and on-board voltage/current regulators, not a single legacy chip. Development is busy (around 930 substantive commits and 248 contributors over five years) and it has gained support for PMICs still being designed into new 2025 hardware, such as NXP's PCA9450 for i.MX 8M boards and PF0900 for i.MX95.
repository signals
sources
- docs.kernel.org
The Linux kernel regulator API is an active generic framework for voltage/current regulators and PMIC-backed supplies, indicating this directory is core infrastructure for many current platforms rather than a single legacy device.
- nxp.jp
NXP still markets the PCA9450 PMIC as active and positions it for current i.MX 8M systems; Linux source in this directory contains REGULATOR_PCA9450 support.
- nxp.com
NXP lists PF09/PF0900 as an active PMIC for i.MX95-class applications; Linux source in this directory contains REGULATOR_PF0900 support.
codex reasoning notes (technical)
This is a real driver directory, but it is a large active subsystem rather than one obsolete chipset family. Upstream health is strong from the supplied static evidence: 930 substantive commits in 5 years, 248 authors, and a most recent substantive touch on 2026-04-07. Local tree inspection via shell showed current Kconfig/Makefile entries for modern PMIC drivers such as MT6360, PCA9450, PF0900, and TPS68470. Lore-first checks were attempted: lore_file_timeline on drivers/regulator/ returned no matches for the directory prefix, lore_regex for removal/deprecation subjects timed out, and lore_nearest was unavailable because the embedding index was not built; with no positive removal signal and very strong recent activity, there is no basis for deprecate/remove. Sources were obtained by web search (kernel docs, NXP product pages).