Re: [PATCH V4 07/26] mm/mmap: Build protect protection_map[] with ARCH_HAS_VM_GET_PAGE_PROT

From: Anshuman Khandual
Date: Fri Jun 24 2022 - 01:37:00 EST




On 6/24/22 10:52, Christophe Leroy wrote:
>
>
> Le 24/06/2022 à 06:43, Anshuman Khandual a écrit :
>> protection_map[] has already been moved inside those platforms which enable
>
> Usually "already" means before your series.
>
> Your series is the one that moves protection_map[] so I would have just
> said "Now that protection_map[] has been moved inside those platforms
> which enable ...."

Got it, will update the commit message.

>
>> ARCH_HAS_VM_GET_PAGE_PROT. Hence generic protection_map[] array now can be
>> protected with CONFIG_ARCH_HAS_VM_GET_PAGE_PROT intead of __P000.
>>
>> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
>> Cc: linux-mm@xxxxxxxxx
>> Cc: linux-kernel@xxxxxxxxxxxxxxx
>> Signed-off-by: Anshuman Khandual <anshuman.khandual@xxxxxxx>
>> ---
>> include/linux/mm.h | 2 +-
>> mm/mmap.c | 5 +----
>> 2 files changed, 2 insertions(+), 5 deletions(-)
>>
>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>> index 237828c2bae2..70d900f6df43 100644
>> --- a/include/linux/mm.h
>> +++ b/include/linux/mm.h
>> @@ -424,7 +424,7 @@ extern unsigned int kobjsize(const void *objp);
>> * mapping from the currently active vm_flags protection bits (the
>> * low four bits) to a page protection mask..
>> */
>> -#ifdef __P000
>> +#ifndef CONFIG_ARCH_HAS_VM_GET_PAGE_PROT
>> extern pgprot_t protection_map[16];
>
> Is this declaration still needed ? I have the feeling that
> protection_map[] is only used in mm/mmap.c now.

At this point generic protection_map[] array is still being used via
this declaration on many (!ARCH_HAS_VM_GET_PAGE_PROT) platforms such
as mips, m68k, arm etc.

>
>> #endif
>>
>> diff --git a/mm/mmap.c b/mm/mmap.c
>> index 55c30aee3999..43db3bd49071 100644
>> --- a/mm/mmap.c
>> +++ b/mm/mmap.c
>> @@ -101,7 +101,7 @@ static void unmap_region(struct mm_struct *mm,
>> * w: (no) no
>> * x: (yes) yes
>> */
>> -#ifdef __P000
>> +#ifndef CONFIG_ARCH_HAS_VM_GET_PAGE_PROT
>> pgprot_t protection_map[16] __ro_after_init = {
>
> Should this be static, as it seems to now be used only in this file ?

This is being used in some platforms as mentioned before.

> And it could also be 'const' instead of __ro_after_init.

Then should be able to be a 'const' wrt mips, m68k, arm platforms.
But should this even be changed, if this is going to be dropped off
eventually ?

>
>> [VM_NONE] = __P000,
>> [VM_READ] = __P001,
>> @@ -120,9 +120,6 @@ pgprot_t protection_map[16] __ro_after_init = {
>> [VM_SHARED | VM_EXEC | VM_WRITE] = __S110,
>> [VM_SHARED | VM_EXEC | VM_WRITE | VM_READ] = __S111
>> };
>> -#endif
>> -
>> -#ifndef CONFIG_ARCH_HAS_VM_GET_PAGE_PROT
>> DECLARE_VM_GET_PAGE_PROT
>> #endif /* CONFIG_ARCH_HAS_VM_GET_PAGE_PROT */
>>