On Fri, May 03, 2019 at 11:05:09AM +0100, Mark Rutland wrote:

Hi,

On Thu, May 02, 2019 at 06:28:31PM +0300, Mike Rapoport wrote:

quoted

The PTE allocations in arm64 are identical to the generic ones modulo the
GFP flags.

Using the generic pte_alloc_one() functions ensures that the user page
tables are allocated with __GFP_ACCOUNT set.

The arm64 definition of PGALLOC_GFP is removed and replaced with
GFP_PGTABLE_USER for p[gum]d_alloc_one() and for KVM memory cache.

The mappings created with create_pgd_mapping() are now using
GFP_PGTABLE_KERNEL.

The conversion to the generic version of pte_free_kernel() removes the NULL
check for pte.

The pte_free() version on arm64 is identical to the generic one and
can be simply dropped.

Signed-off-by: Mike Rapoport <redacted>
---
 arch/arm64/include/asm/pgalloc.h | 43 ++++------------------------------------
 arch/arm64/mm/mmu.c              |  2 +-
 arch/arm64/mm/pgd.c              |  4 ++--
 virt/kvm/arm/mmu.c               |  2 +-
 4 files changed, 8 insertions(+), 43 deletions(-)

[...]

quoted

diff --git a/arch/arm64/mm/pgd.c b/arch/arm64/mm/pgd.c
index 289f911..2ef1a53 100644
--- a/arch/arm64/mm/pgd.c
+++ b/arch/arm64/mm/pgd.c

@@ -31,9 +31,9 @@ static struct kmem_cache *pgd_cache __ro_after_init;
 pgd_t *pgd_alloc(struct mm_struct *mm)
 {
 	if (PGD_SIZE == PAGE_SIZE)
-		return (pgd_t *)__get_free_page(PGALLOC_GFP);
+		return (pgd_t *)__get_free_page(GFP_PGTABLE_USER);
 	else
-		return kmem_cache_alloc(pgd_cache, PGALLOC_GFP);
+		return kmem_cache_alloc(pgd_cache, GFP_PGTABLE_USER);
 }

In efi_virtmap_init() we use pgd_alloc() to allocate a pgd for EFI
runtime services, which we map with a special kernel page table.

I'm not sure if accounting that is problematic, as it's allocated in a
kernel thread off the back of an early_initcall.

The accounting bypasses kernel threads so there should be no problem.
 

Just to check, Is that sound, or do we need a pgd_alloc_kernel()?

Thanks,
Mark.

-- 
Sincerely yours,
Mike.