The quilt patch titled
Subject: mm: refactor do_fault_around()
has been removed from the -mm tree. Its filename was
mm-refactor-do_fault_around.patch
This patch was dropped because it was merged into the mm-stable branch
of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
------------------------------------------------------
From: Lorenzo Stoakes <lstoakes@xxxxxxxxx>
Subject: mm: refactor do_fault_around()
Date: Fri, 17 Mar 2023 21:58:25 +0000
Patch series "Refactor do_fault_around()"
Refactor do_fault_around() to avoid bitwise tricks and rather difficult to
follow logic. Additionally, prefer fault_around_pages to
fault_around_bytes as the operations are performed at a base page
granularity.
This patch (of 2):
The existing logic is confusing and fails to abstract a number of bitwise
tricks.
Use ALIGN_DOWN() to perform alignment, pte_index() to obtain a PTE index
and represent the address range using PTE offsets, which naturally make it
clear that the operation is intended to occur within only a single PTE and
prevent spanning of more than one page table.
We rely on the fact that fault_around_bytes will always be page-aligned,
at least one page in size, a power of two and that it will not exceed
PAGE_SIZE * PTRS_PER_PTE in size (i.e. the address space mapped by a
PTE). These are all guaranteed by fault_around_bytes_set().
Link: https://lkml.kernel.org/r/cover.1679089214.git.lstoakes@xxxxxxxxx
Link: https://lkml.kernel.org/r/d125db1c3665a63b80cea29d56407825482e2262.1679089214.git.lstoakes@xxxxxxxxx
Signed-off-by: Lorenzo Stoakes <lstoakes@xxxxxxxxx>
Cc: David Hildenbrand <david@xxxxxxxxxx>
Cc: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
Cc: Vlastimil Babka <vbabka@xxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/memory.c | 42 ++++++++++++++++++------------------------
1 file changed, 18 insertions(+), 24 deletions(-)
--- a/mm/memory.c~mm-refactor-do_fault_around
+++ a/mm/memory.c
@@ -4449,8 +4449,8 @@ late_initcall(fault_around_debugfs);
* It uses vm_ops->map_pages() to map the pages, which skips the page if it's
* not ready to be mapped: not up-to-date, locked, etc.
*
- * This function doesn't cross the VMA boundaries, in order to call map_pages()
- * only once.
+ * This function doesn't cross VMA or page table boundaries, in order to call
+ * map_pages() and acquire a PTE lock only once.
*
* fault_around_bytes defines how many bytes we'll try to map.
* do_fault_around() expects it to be set to a power of two less than or equal
@@ -4463,27 +4463,19 @@ late_initcall(fault_around_debugfs);
*/
static vm_fault_t do_fault_around(struct vm_fault *vmf)
{
- unsigned long address = vmf->address, nr_pages, mask;
- pgoff_t start_pgoff = vmf->pgoff;
- pgoff_t end_pgoff;
- int off;
-
- nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
- mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;
-
- address = max(address & mask, vmf->vma->vm_start);
- off = ((vmf->address - address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
- start_pgoff -= off;
-
- /*
- * end_pgoff is either the end of the page table, the end of
- * the vma or nr_pages from start_pgoff, depending what is nearest.
- */
- end_pgoff = start_pgoff -
- ((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
- PTRS_PER_PTE - 1;
- end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
- start_pgoff + nr_pages - 1);
+ pgoff_t nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
+ pgoff_t pte_off = pte_index(vmf->address);
+ /* The page offset of vmf->address within the VMA. */
+ pgoff_t vma_off = vmf->pgoff - vmf->vma->vm_pgoff;
+ pgoff_t from_pte, to_pte;
+
+ /* The PTE offset of the start address, clamped to the VMA. */
+ from_pte = max(ALIGN_DOWN(pte_off, nr_pages),
+ pte_off - min(pte_off, vma_off));
+
+ /* The PTE offset of the end address, clamped to the VMA and PTE. */
+ to_pte = min3(from_pte + nr_pages, (pgoff_t)PTRS_PER_PTE,
+ pte_off + vma_pages(vmf->vma) - vma_off) - 1;
if (pmd_none(*vmf->pmd)) {
vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm);
@@ -4491,7 +4483,9 @@ static vm_fault_t do_fault_around(struct
return VM_FAULT_OOM;
}
- return vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
+ return vmf->vma->vm_ops->map_pages(vmf,
+ vmf->pgoff + from_pte - pte_off,
+ vmf->pgoff + to_pte - pte_off);
}
/* Return true if we should do read fault-around, false otherwise */
_
Patches currently in -mm which might be from lstoakes@xxxxxxxxx are
