On 11/22/23 2:53 PM, Vishal Chourasia wrote:
> Changed HOTPLUG_CPU dependency to SMP and either ARCH_HIBERNATION_POSSIBLE or
> ARCH_SUSPEND_POSSIBLE, aligning with systems' suspend/hibernation
> capabilities.
> This update links CPU hotplugging more logically with platforms' capabilities.
>
> Other changes include
>
> 1. configs ARCH_HIBERNATION_POSSIBLE and ARCH_SUSPEND_POSSIBLE are now
> selected
> only if required platform dependencies are met.
>
> 2. Defaults for configs ARCH_HIBERNATION_POSSIBLE and
> ARCH_SUSPEND_POSSIBLE has been set to 'N'
>
> Reported-by: Srikar Dronamraju <sri...@linux.vnet.ibm.com>
> Suggested-by: Aneesh Kumar K V <aneesh.ku...@linux.ibm.com>
> Suggested-by: Michael Ellerman <m...@ellerman.id.au>
> Signed-off-by: Vishal Chourasia <vish...@linux.ibm.com>
>
> v1: https://lore.kernel.org/all/20231114082046.6018-1-vish...@linux.ibm.com
> ---
> During the configuration process with 'make randconfig' followed by
> 'make olddefconfig', we observed a warning indicating an unmet direct
> dependency for the HOTPLUG_CPU option. The dependency in question relates to
> various PowerPC configurations (PPC_PSERIES, PPC_PMAC, PPC_POWERNV,
> FSL_SOC_BOOKE) which were not enabled, yet the HOTPLUG_CPU was being
> erroneously selected due to an implicit assumption by the PM_SLEEP_SMP option.
> This misalignment in dependencies could potentially lead to inconsistent
> kernel
> configuration states, especially when considering the necessary hardware
> support for CPU hot-plugging on PowerPC platforms. The patch aims to correct
> this by ensuring that ARCH_HIBERNATION_POSSIBLE is contingent upon the
> appropriate PowerPC configurations being active.
>
> steps to reproduce (before applying the patch):
>
> Run 'make pseries_le_defconfig'
> Run 'make menuconfig'
> Enable hibernation [ Kernel options -> Hibernation (aka 'suspend to disk') ]
> Disable [ Platform support -> IBM PowerNV (Non-Virtualized) platform support ]
> Disable [ Platform support -> IBM pSeries & new (POWER5-based) iSeries ]
> Enable SMP [ Processor support -> Symmetric multi-processing support ]
> Save the config
> Run 'make olddefconfig'
>
> arch/powerpc/Kconfig | 13 +++++--------
> 1 file changed, 5 insertions(+), 8 deletions(-)
>
> diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
> index 6f105ee4f3cf..6e7e9af2f0c1 100644
> --- a/arch/powerpc/Kconfig
> +++ b/arch/powerpc/Kconfig
> @@ -155,6 +155,8 @@ config PPC
> select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
> select ARCH_HAS_UACCESS_FLUSHCACHE
> select ARCH_HAS_UBSAN_SANITIZE_ALL
> + select ARCH_HIBERNATION_POSSIBLE if (PPC_PSERIES || PPC_PMAC ||
> PPC_POWERNV || FSL_SOC_BOOKE)
> + select ARCH_SUSPEND_POSSIBLE if (ADB_PMU || PPC_EFIKA ||
> PPC_LITE5200 || PPC_83xx || (PPC_85xx && !PPC_E500MC) || PPC_86xx ||
> PPC_PSERIES || 44x || 40x)
We should keep that sorted as explained in comment around that.
> select ARCH_HAVE_NMI_SAFE_CMPXCHG
> select ARCH_KEEP_MEMBLOCK
> select ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE if PPC_RADIX_MMU
> @@ -380,14 +382,10 @@ config DEFAULT_UIMAGE
> Used to allow a board to specify it wants a uImage built by default
>
> config ARCH_HIBERNATION_POSSIBLE
> - bool
> - default y
> + def_bool n
We should be able to keep this
config ARCH_HIBERNATION_POSSIBLE
bool
That is how we have rest of the ARCH_* config. I am not sure we need to
explicitly say "def_bool n"
>
> config ARCH_SUSPEND_POSSIBLE
> - def_bool y
> - depends on ADB_PMU || PPC_EFIKA || PPC_LITE5200 || PPC_83xx || \
> - (PPC_85xx && !PPC_E500MC) || PPC_86xx || PPC_PSERIES \
> - || 44x || 40x
> + def_bool n
>
> config ARCH_SUSPEND_NONZERO_CPU
> def_bool y
> @@ -568,8 +566,7 @@ config ARCH_USING_PATCHABLE_FUNCTION_ENTRY
>
> config HOTPLUG_CPU
> bool "Support for enabling/disabling CPUs"
> - depends on SMP && (PPC_PSERIES || \
> - PPC_PMAC || PPC_POWERNV || FSL_SOC_BOOKE)
> + depends on SMP && (ARCH_HIBERNATION_POSSIBLE || ARCH_SUSPEND_POSSIBLE)
> help
> Say Y here to be able to disable and re-enable individual
> CPUs at runtime on SMP machines.
