Re: [PATCH] mm/sparse: keep mem_section_usage_size() internal

Muchun Song <[email protected]> Tue, 4 Aug 2026 19:10:50 +0800
Newsgroups gmane.linux.kernel,gmane.linux.kernel.mm
Message-ID <[email protected]>

> On Aug 4, 2026, at 18:09, David Hildenbrand (Arm) <[email protected]> =
wrote:
>=20
> On 8/4/26 12:01, Muchun Song wrote:
>> mem_section_usage_size() is only needed by sparsemem implementation =
code,
>> but declaring it in mmzone.h makes the helper visible to all mmzone.h
>> users.
>=20
> Worth mentioning that
>=20
> commit ae751d567baa08342e5e34b378b72a6f9b2cfada
> Author: David Hildenbrand (Arm) <[email protected]>
> Date:   Mon May 11 16:05:34 2026 +0200
>=20
>    mm/bootmem_info: stop marking mem_section_usage as MIX_SECTION_INFO
>=20
> Remove the last external user.

Will do.

>=20
>>=20
>> Move the helper to sparse.h so sparse.c and sparse-vmemmap.c can =
share
>> it through the internal header. While doing so, calculate the =
allocation
>> size with struct_size(), which ties the expression to the =
pageblock_flags
>> trailing array instead of open-coding the struct header plus bitmap =
size.
>>=20
>> Signed-off-by: Muchun Song <[email protected]>
>> ---
>> include/linux/mmzone.h |  1 -
>> mm/sparse.c            | 10 ----------
>> mm/sparse.h            |  6 ++++++
>> 3 files changed, 6 insertions(+), 11 deletions(-)
>>=20
>> diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
>> index a26c8b855222..86891163e0ef 100644
>> --- a/include/linux/mmzone.h
>> +++ b/include/linux/mmzone.h
>> @@ -2070,7 +2070,6 @@ static inline struct mem_section =
*__nr_to_section(unsigned long nr)
>> #endif
>> return &mem_section[root][nr & SECTION_ROOT_MASK];
>> }
>> -extern size_t mem_section_usage_size(void);
>>=20
>> /*
>>  * We use the lower bits of the mem_map pointer to store a little bit =
of
>> diff --git a/mm/sparse.c b/mm/sparse.c
>> index 704a9dec2b9a..46458e5300e2 100644
>> --- a/mm/sparse.c
>> +++ b/mm/sparse.c
>> @@ -213,16 +213,6 @@ static void __init memblocks_present(void)
>> memory_present(nid, start, end);
>> }
>>=20
>> -static unsigned long usemap_size(void)
>> -{
>> - 	return BITS_TO_LONGS(SECTION_BLOCKFLAGS_BITS) * sizeof(unsigned =
long);
>> -}
>> -
>> -size_t mem_section_usage_size(void)
>> -{
>> - 	return sizeof(struct mem_section_usage) + usemap_size();
>> -}
>> -
>> #ifdef CONFIG_SPARSEMEM_VMEMMAP
>> unsigned long __init section_map_size(void)
>> {
>> diff --git a/mm/sparse.h b/mm/sparse.h
>> index 95aa031213f2..4960a17b706f 100644
>> --- a/mm/sparse.h
>> +++ b/mm/sparse.h
>> @@ -47,6 +47,12 @@ static inline void __section_mark_present(struct =
mem_section *ms,
>>=20
>> 	ms->section_mem_map |=3D SECTION_MARKED_PRESENT;
>> }
>> +
>> +static inline size_t mem_section_usage_size(void)
>> +{
>> + 	return struct_size((struct mem_section_usage *)NULL, =
pageblock_flags,
>> +			   BITS_TO_LONGS(SECTION_BLOCKFLAGS_BITS));
>> +}
>=20
> There is struct_size_t() that avoids the manual "NULL" trick.

I hadn't noticed that API before. Thanks for the heads-up.

I'll send a v2 to fix this.

Muchun,
Thanks.

>=20
> --=20
> Cheers,
>=20
> David