Re: [PATCH v2 00/13] mm: remove __ARCH_HAS_5LEVEL_HACK

[Christophe Leroy <christophe.leroy@c-s.fr>]

Le 16/02/2020 à 09:22, Russell King - ARM Linux admin a écrit :
> On Sun, Feb 16, 2020 at 10:18:30AM +0200, Mike Rapoport wrote:
>> From: Mike Rapoport <rppt@linux.ibm.com>
>>
>> Hi,
>>
>> These patches convert several architectures to use page table folding and
>> remove __ARCH_HAS_5LEVEL_HACK along with include/asm-generic/5level-fixup.h.
>>
>> The changes are mostly about mechanical replacement of pgd accessors with p4d
>> ones and the addition of higher levels to page table traversals.
>>
>> All the patches were sent separately to the respective arch lists and
>> maintainers hence the "v2" prefix.
> 
> You fail to explain why this change which adds 488 additional lines of
> code is desirable.
> 

The purpose of the series, ie droping a HACK, is worth it.

However looking at the powerpc patch I have the feeling that this series 
goes behind its purpose.

The number additional lines could be deeply reduced I think if we limit 
the patches to the strict minimum, ie just do things like below instead 
of adding lots of handling of useless levels.

Instead of doing things like:

-	pud = NULL;
+	p4d = NULL;
  	if (pgd_present(*pgd))
-		pud = pud_offset(pgd, gpa);
+		p4d = p4d_offset(pgd, gpa);
+	else
+		new_p4d = p4d_alloc_one(kvm->mm, gpa);
+
+	pud = NULL;
+	if (p4d_present(*p4d))
+		pud = pud_offset(p4d, gpa);
  	else
  		new_pud = pud_alloc_one(kvm->mm, gpa);

It could be limited to:

  	if (pgd_present(*pgd))
-		pud = pud_offset(pgd, gpa);
+		pud = pud_offset(p4d_offset(pgd, gpa), gpa);
  	else
  		new_pud = pud_alloc_one(kvm->mm, gpa);


Christophe