Re: [PATCH] mm/execmem: Make ARCH_WANTS_EXECMEM_LATE depend on EXECMEM

[Anshuman Khandual <anshuman.khandual@arm.com>]

On 12/9/24 22:39, Catalin Marinas wrote:
> On Mon, Dec 09, 2024 at 07:43:32PM +0530, Anshuman Khandual wrote:
>> On 12/9/24 18:06, kernel test robot wrote:
>>> kernel test robot noticed the following build warnings:
>>>
>>> [auto build test WARNING on akpm-mm/mm-everything]
>>>
>>> url:    https://github.com/intel-lab-lkp/linux/commits/Anshuman-Khandual/mm-execmem-Make-ARCH_WANTS_EXECMEM_LATE-depend-on-EXECMEM/20241209-111533
>>> base:   https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything
>>> patch link:    https://lore.kernel.org/r/20241209031251.515903-1-anshuman.khandual%40arm.com
>>> patch subject: [PATCH] mm/execmem: Make ARCH_WANTS_EXECMEM_LATE depend on EXECMEM
>>> config: arm64-kismet-CONFIG_ARCH_WANTS_EXECMEM_LATE-CONFIG_ARM64-0-0 (https://download.01.org/0day-ci/archive/20241209/202412092048.tTzJ5szH-lkp@intel.com/config)
>>> reproduce: (https://download.01.org/0day-ci/archive/20241209/202412092048.tTzJ5szH-lkp@intel.com/reproduce)
>>>
>>> If you fix the issue in a separate patch/commit (i.e. not just a new version of
>>> the same patch/commit), kindly add following tags
>>> | Reported-by: kernel test robot <lkp@intel.com>
>>> | Closes: https://lore.kernel.org/oe-kbuild-all/202412092048.tTzJ5szH-lkp@intel.com/
>>>
>>> kismet warnings: (new ones prefixed by >>)
>>>>> kismet: WARNING: unmet direct dependencies detected for ARCH_WANTS_EXECMEM_LATE when selected by ARM64
>>>    WARNING: unmet direct dependencies detected for ARCH_WANTS_EXECMEM_LATE
>>>      Depends on [n]: EXECMEM [=n]
>>>      Selected by [y]:
>>>      - ARM64 [=y]
>>
>> Although above mentioned config here has CONFIG_ARCH_WANTS_EXECMEM_LATE=y
>> and CONFIG_EXECMEM=n (actually absent), could that really happen though ?
>>
>> Tried making CONFIG_EXECMEM=n (rather deselecting) but that did now allow
>> setting CONFIG_EXECMEM=n, it always remained set. Reverting this patch i.e
>> the current behaviour is also the same (CONFIG_EXECMEM always remains set).
>>
>> select ARCH_WANTS_EXECMEM_LATE if EXECMEM
>>
>> Above statement seems to be selecting EXECMEM as well but the intent seems
>> to be a dependency check instead! OR arch selection for a config brings in
>> all its dependencies as well.
>>
>> If the selection is always, should this change be like the following ?
>>
>> config ARCH_WANTS_EXECMEM_LATE
>>  	bool
>> 	selects EXECMEM
> 
> This is wrong. What ARCH_WANTS_EXECMEM_LATE means is that, _if_ EXECMEM
> is enabled, the arch code require its late initialisation. Currently for

Which implies that even if ARCH_WANTS_EXECMEM_LATE is only applicable when
EXECMEM is enabled, there is no real dependence on the later for platforms
that need a certain init behaviour.

> arm64 EXECMEM is selected by KPROBES, BPF_JIT and MODULES. So if you

EXECMEM gets enabled from the above, not via ARCH_WANTS_EXECMEM_LATE as I
had anticipated earlier.

> don't have any of these on, you don't need EXECMEM either, no point in
> selecting it above just because arm64 wants a specific behaviour (late
> initialisation) for EXECMEM.

Right, specific behaviour preference for a function on a platform should
not necessarily pull in that function itself.

> 
> If you do want a patch for this, I think for arm64 we simply need to
> select ARCH_WANTS_EXECMEM_LATE unconditionally. Not much use in having
> any dependencies really, it just adds to the clutter. And don't change
> the core Kconfig either.

Sure, will do.