Re: [PATCH v2 2/3] hugetlb: Convert hugetlb_no_page() to use struct vm_fault

[Vishal Moola <vishal.moola@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2169 bytes --]

On Sun, Apr 07, 2024 at 04:59:13PM +0800, Muchun Song wrote:
> 
> 
> > On Apr 5, 2024, at 03:58, Vishal Moola <vishal.moola@gmail.com> wrote:
> > 
> > On Thu, Apr 4, 2024 at 5:49 AM Oscar Salvador <osalvador@suse.de> wrote:
> >> 
> >> On Mon, Apr 01, 2024 at 01:26:50PM -0700, Vishal Moola (Oracle) wrote:
> >>> hugetlb_no_page() can use the struct vm_fault passed in from
> >>> hugetlb_fault(). This alleviates the stack by consolidating 7
> >>> variables into a single struct.
> >>> 
> >>> Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
> >>> ---
> >>> mm/hugetlb.c | 59 ++++++++++++++++++++++++++--------------------------
> >>> 1 file changed, 29 insertions(+), 30 deletions(-)
> >>> 
> >>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> >>> index 360b82374a89..aca2f11b4138 100644
> >>> --- a/mm/hugetlb.c
> >>> +++ b/mm/hugetlb.c
> >>> @@ -6189,9 +6189,7 @@ static bool hugetlb_pte_stable(struct hstate *h, struct mm_struct *mm,
> >>> 
> >>> static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
> >>>                      struct vm_area_struct *vma,
> >>> -                     struct address_space *mapping, pgoff_t idx,
> >>> -                     unsigned long address, pte_t *ptep,
> >>> -                     pte_t old_pte, unsigned int flags,
> >>> +                     struct address_space *mapping,
> >> 
> >> AFAICS all this can be self-contained in vm_fault struct.
> >> vmf->vma->mm and vmf->vma.
> >> I mean, if we want to convert this interface, why not going all the way?
> >> 
> >> Looks a bit odd some fields yes while some others remain.
> >> 
> >> Or am I missing something?
> > 
> > Mainly just minimizing code churn, we would either unnecessarily
> > change multiple lines using vma or have to declare the variables
> > again anyways (or have extra churn I didn't like).
> 
> I don't think adding some variables is a problem. I suppose the compiler
> could do some optimization for us. So I think it is better to pass
> only one argument vmf to hugetlb_no_page(). Otherwise, LGTM.

Alright we can get rid of the vm_area_struct and mm_struct arguments as
well.

Andrew, could you please fold the attached patch into this one? 

[-- Attachment #2: 0001-hugetlb-Simplify-hugetlb_no_page-arguments.patch --]
[-- Type: text/plain, Size: 1513 bytes --]

From 891e085115a06f638e238bea267d520bb2432fba Mon Sep 17 00:00:00 2001
From: "Vishal Moola (Oracle)" <vishal.moola@gmail.com>
Date: Mon, 8 Apr 2024 10:17:54 -0700
Subject: [PATCH 1/2] hugetlb: Simplify hugetlb_no_page() arguments

To simplify the function arguments, as suggested by Oscar and Muchun.

Suggested-by: Muchun Song <muchun.song@linux.dev>
Suggested-by: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
---
 mm/hugetlb.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 456c81fbf8f5..05fe610f4699 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -6186,11 +6186,11 @@ static bool hugetlb_pte_stable(struct hstate *h, struct mm_struct *mm,
 	return same;
 }
 
-static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
-			struct vm_area_struct *vma,
-			struct address_space *mapping,
+static vm_fault_t hugetlb_no_page(struct address_space *mapping,
 			struct vm_fault *vmf)
 {
+	struct vm_area_struct *vma = vmf->vma;
+	struct mm_struct *mm = vma->vm_mm;
 	struct hstate *h = hstate_vma(vma);
 	vm_fault_t ret = VM_FAULT_SIGBUS;
 	int anon_rmap = 0;
@@ -6483,7 +6483,7 @@ vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 		 * hugetlb_no_page will drop vma lock and hugetlb fault
 		 * mutex internally, which make us return immediately.
 		 */
-		return hugetlb_no_page(mm, vma, mapping, &vmf);
+		return hugetlb_no_page(mapping, &vmf);
 	}
 
 	ret = 0;
-- 
2.43.0