clk: mmp: allow COMPILE_TEST builds
Subsystems (2)
PATCH 1/1 clk: mmp: allow COMPILE_TEST builds
The MMP clock directory is only entered for ARCH_MMP, even though the PXA1908 and MMP2 audio clock symbols can already be selected for COMPILE_TEST.
Add a hidden MMP clock family gate, select it from the compile-testable MMP clock symbols, and use that symbol for the parent Makefile descent. Keep ARCH_MMP as the default provider for existing platform builds.
Tested with:
make LLVM=1 ARCH=loongarch drivers/clk/mmp/
--- drivers/clk/Kconfig | 2 ++ drivers/clk/Makefile | 2 +- drivers/clk/mmp/Kconfig | 5 +++++ 3 files changed, 8 insertions(+), 1 deletion(-)
drivers/clk/Kconfig +2 -0
@@ -457,12 +457,14 @@ config COMMON_CLK_STM32H7 config COMMON_CLK_MMP2 def_bool COMMON_CLK && (MACH_MMP2_DT || MACH_MMP3_DT) + select COMMON_CLK_MMP help Support for Marvell MMP2 and MMP3 SoC clocks config COMMON_CLK_MMP2_AUDIO tristate "Clock driver for MMP2 Audio subsystem" depends on COMMON_CLK_MMP2 || COMPILE_TEST + select COMMON_CLK_MMP help This driver supports clocks for Audio subsystem on MMP2 SoC.
drivers/clk/Makefile +1 -1
@@ -130,7 +130,7 @@ obj-y += mediatek/ obj-$(CONFIG_ARCH_MESON) += meson/ obj-y += microchip/ ifeq ($(CONFIG_COMMON_CLK), y) -obj-$(CONFIG_ARCH_MMP) += mmp/ +obj-$(CONFIG_COMMON_CLK_MMP) += mmp/ endif obj-y += mstar/ obj-y += mvebu/
drivers/clk/mmp/Kconfig +5 -0
@@ -1,10 +1,15 @@ # SPDX-License-Identifier: GPL-2.0-only +config COMMON_CLK_MMP + bool + default ARCH_MMP + config COMMON_CLK_PXA1908 bool "Clock driver for Marvell PXA1908" depends on ARCH_MMP || COMPILE_TEST depends on OF default y if ARCH_MMP && ARM64 + select COMMON_CLK_MMP select AUXILIARY_BUS help This driver supports the Marvell PXA1908 SoC clocks.
Signature · 1 line
2.55.0
On Sunday, 19 July 2026 23:43:49 Central European Summer Time Rosen Penev wrote:
[...]
drivers/clk/Makefile obj-y +=
Quoted 5 lines
drivers/clk/Makefile +0
-0@@ -130,7 +130,7 @@ obj-y +=
mediatek/
drivers/clk/Makefile obj-y += mediatek/ · line 133
Quoted 5 lines
drivers/clk/Makefile +1
-1obj-$(CONFIG_ARCH_MESON) += meson/ obj-y += microchip/ ifeq ($(CONFIG_COMMON_CLK), y) -obj-$(CONFIG_ARCH_MMP) += mmp/ +obj-$(CONFIG_COMMON_CLK_MMP) += mmp/
Could it be possible to select mmp unconditionally, like most of the other subtrees are? AFAICT, that would also require creating new Kconfig symbols for the PXA168, PXA910 and PXA1928 clocks, but I suspect it may be the more proper choice in the long term.
Regards,
--
Duje
Hi Duje,
Quoted 11 lines · Duje Mihanović
On Sunday, 19 July 2026 23:43:49 Central European Summer Time Rosen Penev wrote:
[...]
drivers/clk/Makefile obj-y +=
Quoted 5 lines
drivers/clk/Makefile +0
-0@@ -130,7 +130,7 @@ obj-y +=mediatek/
Quoted 10 lines
drivers/clk/Makefile obj-y += mediatek/ · line 133
Quoted 5 lines
drivers/clk/Makefile +1
-1obj-$(CONFIG_ARCH_MESON) += meson/ obj-y += microchip/ ifeq ($(CONFIG_COMMON_CLK), y) -obj-$(CONFIG_ARCH_MMP) += mmp/ +obj-$(CONFIG_COMMON_CLK_MMP) += mmp/Could it be possible to select mmp unconditionally, like most of the other subtrees are? AFAICT, that would also require creating new
Kconfigsymbols for the PXA168, PXA910 and PXA1928 clocks, but I suspect it may be the more proper choice in the long term.
Just by sheer numbers, the CONFIG_COMMON_CLK_XXX approach is the most common use case right now.
x1:~/src/linux/linux-next ((next-20260720) %)$ grep "CONFIG_COMMON_CLK" drivers/clk/Makefile | grep "+=" | wc -l
68
x1:~/src/linux/linux-next ((next-20260720) %)$ grep "obj-y" drivers/clk/Makefile | wc -l
25I have no strong opinions about this either way other than to drive consistency across various drivers as much as possible.
Currently the mmp directory currently only has COMMON_CLK_PXA1908 defined. When the other PXA* drivers are introduced, will there be a CONFIG_COMMON_CLK_MMA_COMMON that all of the drivers will use?
If so, we could put that MMA_COMMON in the toplevel clk Makefile.
If not, then yes I agree that obj-y will be fine.
Brian
Quoted 26 lines · Brian Masney
Hi Duje,
Quoted 11 lines · Duje Mihanović
On Sunday, 19 July 2026 23:43:49 Central European Summer Time Rosen Penev wrote:
[...]
drivers/clk/Makefile obj-y +=
Quoted 5 lines
drivers/clk/Makefile +0
-0@@ -130,7 +130,7 @@ obj-y +=mediatek/
Quoted 10 lines
drivers/clk/Makefile obj-y += mediatek/ · line 133
Quoted 5 lines
drivers/clk/Makefile +1
-1obj-$(CONFIG_ARCH_MESON) += meson/ obj-y += microchip/ ifeq ($(CONFIG_COMMON_CLK), y) -obj-$(CONFIG_ARCH_MMP) += mmp/ +obj-$(CONFIG_COMMON_CLK_MMP) += mmp/Could it be possible to select mmp unconditionally, like most of the other subtrees are? AFAICT, that would also require creating new
Kconfigsymbols for the PXA168, PXA910 and PXA1928 clocks, but I suspect it may be the more proper choice in the long term.
Quoted 2 lines
Just by sheer numbers, the
CONFIG_COMMON_CLK_XXXapproach is the most common use case right now.
It's much simpler this way.
Quoted 19 lines
x1:~/src/linux/linux-next ((next-20260720) %)$ grep "CONFIG_COMMON_CLK" drivers/clk/Makefile | grep "+=" | wc -l 68 x1:~/src/linux/linux-next ((next-20260720) %)$ grep "obj-y" drivers/clk/Makefile | wc -l 25I have no strong opinions about this either way other than to drive consistency across various drivers as much as possible.
Currently the mmp directory currently only has
COMMON_CLK_PXA1908defined. When the other PXA* drivers are introduced, will there be aCONFIG_COMMON_CLK_MMA_COMMONthat all of the drivers will use?If so, we could put that
MMA_COMMONin the toplevel clkMakefile.If not, then yes I agree that obj-y will be fine.
Brian
Brian Masney
AppliedQuoted 1 line · Rosen Penev
clk: mmp: allow
COMPILE_TESTbuilds
Applied, thanks!
[1/1] clk: mmp: allow COMPILE_TEST builds
commit: d0b72c88671b3c46342bbebc9b90280890436c98
Best regards,
Signature · 1 line
Brian Masney <[email protected]>