add led

hc

2024-05-10 10ebd8556b7990499c896a550e3d416b444211e6

 kernel/include/linux/userfaultfd_k.h
..
..
@@ -14,6 +14,11 @@
14
14
 #include <linux/userfaultfd.h> /* linux/include/uapi/linux/userfaultfd.h */
15
15
 
16
16
 #include <linux/fcntl.h>
17
+#include <linux/mm.h>
18
+#include <asm-generic/pgtable_uffd.h>
19
+
20
+/* The set of all possible UFFD-related VM flags. */
21
+#define __VM_UFFD_FLAGS (VM_UFFD_MISSING | VM_UFFD_WP | VM_UFFD_MINOR)
17
22
 
18
23
 /*
19
24
  * CAREFUL: Check include/uapi/asm-generic/fcntl.h when defining
..
..
@@ -28,15 +33,43 @@
28
33
 #define UFFD_SHARED_FCNTL_FLAGS (O_CLOEXEC | O_NONBLOCK)
29
34
 #define UFFD_FLAGS_SET (EFD_SHARED_FCNTL_FLAGS)
30
35
 
36
+extern int sysctl_unprivileged_userfaultfd;
37
+
31
38
 extern vm_fault_t handle_userfault(struct vm_fault *vmf, unsigned long reason);
39
+
40
+/*
41
+ * The mode of operation for __mcopy_atomic and its helpers.
42
+ *
43
+ * This is almost an implementation detail (mcopy_atomic below doesn't take this
44
+ * as a parameter), but it's exposed here because memory-kind-specific
45
+ * implementations (e.g. hugetlbfs) need to know the mode of operation.
46
+ */
47
+enum mcopy_atomic_mode {
48
+	/* A normal copy_from_user into the destination range. */
49
+	MCOPY_ATOMIC_NORMAL,
50
+	/* Don't copy; map the destination range to the zero page. */
51
+	MCOPY_ATOMIC_ZEROPAGE,
52
+	/* Just install pte(s) with the existing page(s) in the page cache. */
53
+	MCOPY_ATOMIC_CONTINUE,
54
+};
55
+
56
+extern int mfill_atomic_install_pte(struct mm_struct *dst_mm, pmd_t *dst_pmd,
57
+				    struct vm_area_struct *dst_vma,
58
+				    unsigned long dst_addr, struct page *page,
59
+				    bool newly_allocated, bool wp_copy);
32
60
 
33
61
 extern ssize_t mcopy_atomic(struct mm_struct *dst_mm, unsigned long dst_start,
34
62
 			    unsigned long src_start, unsigned long len,
35
-			    bool *mmap_changing);
63
+			    bool *mmap_changing, __u64 mode);
36
64
 extern ssize_t mfill_zeropage(struct mm_struct *dst_mm,
37
65
 			      unsigned long dst_start,
38
66
 			      unsigned long len,
39
67
 			      bool *mmap_changing);
68
+extern ssize_t mcopy_continue(struct mm_struct *dst_mm, unsigned long dst_start,
69
+			      unsigned long len, bool *mmap_changing);
70
+extern int mwriteprotect_range(struct mm_struct *dst_mm,
71
+			       unsigned long start, unsigned long len,
72
+			       bool enable_wp, bool *mmap_changing);
40
73
 
41
74
 /* mm helpers */
42
75
 static inline bool is_mergeable_vm_userfaultfd_ctx(struct vm_area_struct *vma,
..
..
@@ -45,14 +78,52 @@
45
78
 	return vma->vm_userfaultfd_ctx.ctx == vm_ctx.ctx;
46
79
 }
47
80
 
81
+/*
82
+ * Never enable huge pmd sharing on some uffd registered vmas:
83
+ *
84
+ * - VM_UFFD_WP VMAs, because write protect information is per pgtable entry.
85
+ *
86
+ * - VM_UFFD_MINOR VMAs, because otherwise we would never get minor faults for
87
+ *   VMAs which share huge pmds. (If you have two mappings to the same
88
+ *   underlying pages, and fault in the non-UFFD-registered one with a write,
89
+ *   with huge pmd sharing this would *also* setup the second UFFD-registered
90
+ *   mapping, and we'd not get minor faults.)
91
+ */
92
+static inline bool uffd_disable_huge_pmd_share(struct vm_area_struct *vma)
93
+{
94
+	return vma->vm_flags & (VM_UFFD_WP | VM_UFFD_MINOR);
95
+}
96
+
48
97
 static inline bool userfaultfd_missing(struct vm_area_struct *vma)
49
98
 {
50
99
 	return vma->vm_flags & VM_UFFD_MISSING;
51
100
 }
52
101
 
102
+static inline bool userfaultfd_wp(struct vm_area_struct *vma)
103
+{
104
+	return vma->vm_flags & VM_UFFD_WP;
105
+}
106
+
107
+static inline bool userfaultfd_minor(struct vm_area_struct *vma)
108
+{
109
+	return vma->vm_flags & VM_UFFD_MINOR;
110
+}
111
+
112
+static inline bool userfaultfd_pte_wp(struct vm_area_struct *vma,
113
+				      pte_t pte)
114
+{
115
+	return userfaultfd_wp(vma) && pte_uffd_wp(pte);
116
+}
117
+
118
+static inline bool userfaultfd_huge_pmd_wp(struct vm_area_struct *vma,
119
+					   pmd_t pmd)
120
+{
121
+	return userfaultfd_wp(vma) && pmd_uffd_wp(pmd);
122
+}
123
+
53
124
 static inline bool userfaultfd_armed(struct vm_area_struct *vma)
54
125
 {
55
-	return vma->vm_flags & (VM_UFFD_MISSING | VM_UFFD_WP);
126
+	return vma->vm_flags & __VM_UFFD_FLAGS;
56
127
 }
57
128
 
58
129
 extern int dup_userfaultfd(struct vm_area_struct *, struct list_head *);
..
..
@@ -94,6 +165,29 @@
94
165
 	return false;
95
166
 }
96
167
 
168
+static inline bool userfaultfd_wp(struct vm_area_struct *vma)
169
+{
170
+	return false;
171
+}
172
+
173
+static inline bool userfaultfd_minor(struct vm_area_struct *vma)
174
+{
175
+	return false;
176
+}
177
+
178
+static inline bool userfaultfd_pte_wp(struct vm_area_struct *vma,
179
+				      pte_t pte)
180
+{
181
+	return false;
182
+}
183
+
184
+static inline bool userfaultfd_huge_pmd_wp(struct vm_area_struct *vma,
185
+					   pmd_t pmd)
186
+{
187
+	return false;
188
+}
189
+
190
+
97
191
 static inline bool userfaultfd_armed(struct vm_area_struct *vma)
98
192
 {
99
193
 	return false;