On 22/09/2023 09:10, Christophe Leroy wrote:

Le 22/09/2023 à 09:33, Ryan Roberts a écrit :

quoted

On 22/09/2023 07:56, Christophe Leroy wrote:

quoted

Le 21/09/2023 à 18:20, Ryan Roberts a écrit :

quoted

In order to fix a bug, arm64 needs access to the vma inside it's
implementation of set_huge_pte_at(). Provide for this by converting the
mm parameter to be a vma. Any implementations that require the mm can
access it via vma->vm_mm.

This commit makes the required powerpc modifications. Separate commits
update the other arches and core code, before the actual bug is fixed in
arm64.

No behavioral changes intended.

Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
---
   arch/powerpc/include/asm/nohash/32/hugetlb-8xx.h | 3 ++-
   arch/powerpc/mm/book3s64/hugetlbpage.c           | 2 +-
   arch/powerpc/mm/book3s64/radix_hugetlbpage.c     | 2 +-
   arch/powerpc/mm/nohash/8xx.c                     | 2 +-
   arch/powerpc/mm/pgtable.c                        | 7 ++++++-
   5 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/arch/powerpc/include/asm/nohash/32/hugetlb-8xx.h b/arch/powerpc/include/asm/nohash/32/hugetlb-8xx.h
index de092b04ee1a..fff8cd726bc7 100644
--- a/arch/powerpc/include/asm/nohash/32/hugetlb-8xx.h
+++ b/arch/powerpc/include/asm/nohash/32/hugetlb-8xx.h

@@ -46,7 +46,8 @@ static inline int check_and_get_huge_psize(int shift)
   }
   
   #define __HAVE_ARCH_HUGE_SET_HUGE_PTE_AT
-void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte);
+void set_huge_pte_at(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep, pte_t pte);
+void __set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte);

Don't add the burden of an additional function, see below

quoted

   
   #define __HAVE_ARCH_HUGE_PTE_CLEAR
   static inline void huge_pte_clear(struct mm_struct *mm, unsigned long addr,

diff --git a/arch/powerpc/mm/book3s64/hugetlbpage.c b/arch/powerpc/mm/book3s64/hugetlbpage.c
index 3bc0eb21b2a0..ae7fd7c90eb8 100644
--- a/arch/powerpc/mm/book3s64/hugetlbpage.c
+++ b/arch/powerpc/mm/book3s64/hugetlbpage.c

@@ -147,7 +147,7 @@ void huge_ptep_modify_prot_commit(struct vm_area_struct *vma, unsigned long addr
   	if (radix_enabled())
   		return radix__huge_ptep_modify_prot_commit(vma, addr, ptep,
   							   old_pte, pte);
-	set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+	set_huge_pte_at(vma, addr, ptep, pte);
   }
   
   void __init hugetlbpage_init_defaultsize(void)

diff --git a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c b/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
index 17075c78d4bc..7cd40a334c3a 100644
--- a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
+++ b/arch/powerpc/mm/book3s64/radix_hugetlbpage.c

@@ -58,5 +58,5 @@ void radix__huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
   	    atomic_read(&mm->context.copros) > 0)
   		radix__flush_hugetlb_page(vma, addr);
   
-	set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+	set_huge_pte_at(vma, addr, ptep, pte);
   }

diff --git a/arch/powerpc/mm/nohash/8xx.c b/arch/powerpc/mm/nohash/8xx.c
index dbbfe897455d..650a7a8496b6 100644
--- a/arch/powerpc/mm/nohash/8xx.c
+++ b/arch/powerpc/mm/nohash/8xx.c

@@ -91,7 +91,7 @@ static int __ref __early_map_kernel_hugepage(unsigned long va, phys_addr_t pa,
   	if (new && WARN_ON(pte_present(*ptep) && pgprot_val(prot)))
   		return -EINVAL;
   
-	set_huge_pte_at(&init_mm, va, ptep, pte_mkhuge(pfn_pte(pa >> PAGE_SHIFT, prot)));
+	__set_huge_pte_at(&init_mm, va, ptep, pte_mkhuge(pfn_pte(pa >> PAGE_SHIFT, prot)));

Call set_huge_pte_at() with a NULL vma instead.

I'm happy to take your proposed approach if that's your preference. Another
option is to use a dummy VMA, as I have done in the core code, for the one call
site that calls set_huge_pte_at() with init_mm:

struct vm_area_struct vma = TLB_FLUSH_VMA(&init_mm, 0);

This is an existing macro that creates a dummy vma with vma->vm_mm filled in.
Then I pass &vma to the function.

I don't like that, I prefer the solution I proposed. We already have a 
couple places where powerpc do things based on whether vma is NULL or not.

quoted

Or yet another option would be to keep the mm param as is in set_huge_pte_at(),
and add a size param to the function. But then all call sites have the burden of
figuring out the size of the huge pte (although I think most know already).

Indeed.

arch_make_huge_pte() used to take a vma until commit 79c1c594f49a 
("mm/hugetlb: change parameters of arch_make_huge_pte()").

Should we try and have the same approach ? Or is it irrelevant ?

See [1]; I'm going to rework to pass mm + size parameter since the current
approach will break riscv.

[1]
https://lore.kernel.org/linux-arm-kernel/fc85f58e-e8ed-4b24-a3e5-d6288156595e@arm.com/ (local)

Christophe

quoted

Thanks,
Ryan

quoted

quoted

   
   	return 0;
   }

diff --git a/arch/powerpc/mm/pgtable.c b/arch/powerpc/mm/pgtable.c
index 3f86fd217690..9cbcb561a4d8 100644
--- a/arch/powerpc/mm/pgtable.c
+++ b/arch/powerpc/mm/pgtable.c

@@ -288,7 +288,7 @@ int huge_ptep_set_access_flags(struct vm_area_struct *vma,
   }
   
   #if defined(CONFIG_PPC_8xx)
-void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
+void __set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)

Keep it as set_huge_pte_at() with vma argument.

quoted

   {
   	pmd_t *pmd = pmd_off(mm, addr);

Change to:

	pmd_t *pmd = vma ? pmd_off(vma->vm_mm, addr) : pmd_off_k(addr);

quoted

   	pte_basic_t val;

@@ -310,6 +310,11 @@ void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_
   	for (i = 0; i < num; i++, entry++, val += SZ_4K)
   		*entry = val;
   }
+
+void set_huge_pte_at(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep, pte_t pte)
+{
+	__set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+}

Remove this burden.

quoted

   #endif
   #endif /* CONFIG_HUGETLB_PAGE */
   

Christophe