[RFC,13/24] mm/memory: add prepare_mm_fault() function

Commit Message

Michel Lespinasse Feb. 24, 2020, 8:30 p.m. UTC

Add a prepare_mm_fault() function, which may allocate an anon_vma if
required for the incoming fault.

This is because the anon_vma must be allocated in the vma of record,
while in the range locked case, the fault will operate on a pseudo-vma.

Signed-off-by: Michel Lespinasse <[e-mail]>
---
 include/linux/mm.h | 14 ++++++++++++++
 mm/memory.c        | 26 ++++++++++++++++++++++++++
 2 files changed, 40 insertions(+)

Patch

diff --git include/linux/mm.h include/linux/mm.h
index 1b6b022064b4..43b7121ae005 100644
--- include/linux/mm.h
+++ include/linux/mm.h
@@ -1460,6 +1460,15 @@  int generic_error_remove_page(struct address_space *mapping, struct page *page);
 int invalidate_inode_page(struct page *page);
 
 #ifdef CONFIG_MMU
+extern vm_fault_t __prepare_mm_fault(struct vm_area_struct *vma,
+		unsigned int flags);
+static inline vm_fault_t prepare_mm_fault(struct vm_area_struct *vma,
+		unsigned int flags)
+{
+	if (likely(vma->anon_vma))
+		return 0;
+	return __prepare_mm_fault(vma, flags);
+}
 extern vm_fault_t handle_mm_fault_range(struct vm_area_struct *vma,
 			unsigned long address, unsigned int flags,
 			struct mm_lock_range *range);
@@ -1477,6 +1486,11 @@  void unmap_mapping_pages(struct address_space *mapping,
 void unmap_mapping_range(struct address_space *mapping,
 		loff_t const holebegin, loff_t const holelen, int even_cows);
 #else
+static inline vm_fault_t prepare_mm_fault(struct vm_area_struct *vma,
+		unsigned int flags)
+{
+	return 0;
+}
 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
 		unsigned long address, unsigned int flags)
 {
diff --git mm/memory.c mm/memory.c
index 3da4ae504957..9d0b761833fe 100644
--- mm/memory.c
+++ mm/memory.c
@@ -4129,6 +4129,32 @@  static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
 	return handle_pte_fault(&vmf);
 }
 
+vm_fault_t __prepare_mm_fault(struct vm_area_struct *vma, unsigned int flags)
+{
+	vm_fault_t ret = 0;
+
+	if (vma_is_anonymous(vma) ||
+	    ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) ||
+	    (is_vm_hugetlb_page(vma) && !(vma->vm_flags & VM_MAYSHARE))) {
+		if (flags & FAULT_FLAG_USER)
+			mem_cgroup_enter_user_fault();
+		if (unlikely(__anon_vma_prepare(vma)))
+			ret = VM_FAULT_OOM;
+		if (flags & FAULT_FLAG_USER) {
+			mem_cgroup_exit_user_fault();
+			/*
+			 * The task may have entered a memcg OOM situation but
+			 * if the allocation error was handled gracefully (no
+			 * VM_FAULT_OOM), there is no need to kill anything.
+			 * Just clean up the OOM state peacefully.
+			 */
+			if (task_in_memcg_oom(current) && !(ret & VM_FAULT_OOM))
+				mem_cgroup_oom_synchronize(false);
+		}
+	}
+	return ret;
+}
+
 /*
  * By the time we get here, we already hold the mm semaphore
  *