tools: add skeleton code for userland testing of VMA logic
authorLorenzo Stoakes <lorenzo.stoakes@oracle.com>
Mon, 29 Jul 2024 11:50:41 +0000 (12:50 +0100)
committerAndrew Morton <akpm@linux-foundation.org>
Mon, 2 Sep 2024 03:25:55 +0000 (20:25 -0700)
Establish a new userland VMA unit testing implementation under
tools/testing which utilises existing logic providing maple tree support
in userland utilising the now-shared code previously exclusive to radix
tree testing.

This provides fundamental VMA operations whose API is defined in mm/vma.h,
while stubbing out superfluous functionality.

This exists as a proof-of-concept, with the test implementation functional
and sufficient to allow userland compilation of vma.c, but containing only
cursory tests to demonstrate basic functionality.

Link: https://lkml.kernel.org/r/533ffa2eec771cbe6b387dd049a7f128a53eb616.1722251717.git.lorenzo.stoakes@oracle.com
Signed-off-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Tested-by: SeongJae Park <sj@kernel.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Brendan Higgins <brendanhiggins@google.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: David Gow <davidgow@google.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Kees Cook <kees@kernel.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Rae Moar <rmoar@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Pengfei Xu <pengfei.xu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
MAINTAINERS
tools/testing/vma/.gitignore [new file with mode: 0644]
tools/testing/vma/Makefile [new file with mode: 0644]
tools/testing/vma/linux/atomic.h [new file with mode: 0644]
tools/testing/vma/linux/mmzone.h [new file with mode: 0644]
tools/testing/vma/vma.c [new file with mode: 0644]
tools/testing/vma/vma_internal.h [new file with mode: 0644]

index 4a7b6e25bc77d21ff15895bcf70531aa84aec25e..41b15083e4a1e7625b23232915007a662d1b6408 100644 (file)
@@ -24423,6 +24423,7 @@ T:      git git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
 F:     mm/vma.c
 F:     mm/vma.h
 F:     mm/vma_internal.h
+F:     tools/testing/vma/
 
 VMALLOC
 M:     Andrew Morton <akpm@linux-foundation.org>
diff --git a/tools/testing/vma/.gitignore b/tools/testing/vma/.gitignore
new file mode 100644 (file)
index 0000000..b003258
--- /dev/null
@@ -0,0 +1,7 @@
+# SPDX-License-Identifier: GPL-2.0-only
+generated/bit-length.h
+generated/map-shift.h
+generated/autoconf.h
+idr.c
+radix-tree.c
+vma
diff --git a/tools/testing/vma/Makefile b/tools/testing/vma/Makefile
new file mode 100644 (file)
index 0000000..bfc905d
--- /dev/null
@@ -0,0 +1,16 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+.PHONY: default
+
+default: vma
+
+include ../shared/shared.mk
+
+OFILES = $(SHARED_OFILES) vma.o maple-shim.o
+TARGETS = vma
+
+vma:   $(OFILES) vma_internal.h ../../../mm/vma.c ../../../mm/vma.h
+       $(CC) $(CFLAGS) -o $@ $(OFILES) $(LDLIBS)
+
+clean:
+       $(RM) $(TARGETS) *.o radix-tree.c idr.c generated/map-shift.h generated/bit-length.h generated/autoconf.h
diff --git a/tools/testing/vma/linux/atomic.h b/tools/testing/vma/linux/atomic.h
new file mode 100644 (file)
index 0000000..e01f66f
--- /dev/null
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#ifndef _LINUX_ATOMIC_H
+#define _LINUX_ATOMIC_H
+
+#define atomic_t int32_t
+#define atomic_inc(x) uatomic_inc(x)
+#define atomic_read(x) uatomic_read(x)
+#define atomic_set(x, y) do {} while (0)
+#define U8_MAX UCHAR_MAX
+
+#endif /* _LINUX_ATOMIC_H */
diff --git a/tools/testing/vma/linux/mmzone.h b/tools/testing/vma/linux/mmzone.h
new file mode 100644 (file)
index 0000000..33cd151
--- /dev/null
@@ -0,0 +1,38 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#ifndef _LINUX_MMZONE_H
+#define _LINUX_MMZONE_H
+
+#include <linux/atomic.h>
+
+struct pglist_data *first_online_pgdat(void);
+struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
+
+#define for_each_online_pgdat(pgdat)                   \
+       for (pgdat = first_online_pgdat();              \
+            pgdat;                                     \
+            pgdat = next_online_pgdat(pgdat))
+
+enum zone_type {
+       __MAX_NR_ZONES
+};
+
+#define MAX_NR_ZONES __MAX_NR_ZONES
+#define MAX_PAGE_ORDER 10
+#define MAX_ORDER_NR_PAGES (1 << MAX_PAGE_ORDER)
+
+#define pageblock_order                MAX_PAGE_ORDER
+#define pageblock_nr_pages     BIT(pageblock_order)
+#define pageblock_align(pfn)   ALIGN((pfn), pageblock_nr_pages)
+#define pageblock_start_pfn(pfn)       ALIGN_DOWN((pfn), pageblock_nr_pages)
+
+struct zone {
+       atomic_long_t           managed_pages;
+};
+
+typedef struct pglist_data {
+       struct zone node_zones[MAX_NR_ZONES];
+
+} pg_data_t;
+
+#endif /* _LINUX_MMZONE_H */
diff --git a/tools/testing/vma/vma.c b/tools/testing/vma/vma.c
new file mode 100644 (file)
index 0000000..48e033c
--- /dev/null
@@ -0,0 +1,207 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "maple-shared.h"
+#include "vma_internal.h"
+
+/*
+ * Directly import the VMA implementation here. Our vma_internal.h wrapper
+ * provides userland-equivalent functionality for everything vma.c uses.
+ */
+#include "../../../mm/vma.c"
+
+const struct vm_operations_struct vma_dummy_vm_ops;
+
+#define ASSERT_TRUE(_expr)                                             \
+       do {                                                            \
+               if (!(_expr)) {                                         \
+                       fprintf(stderr,                                 \
+                               "Assert FAILED at %s:%d:%s(): %s is FALSE.\n", \
+                               __FILE__, __LINE__, __FUNCTION__, #_expr); \
+                       return false;                                   \
+               }                                                       \
+       } while (0)
+#define ASSERT_FALSE(_expr) ASSERT_TRUE(!(_expr))
+#define ASSERT_EQ(_val1, _val2) ASSERT_TRUE((_val1) == (_val2))
+#define ASSERT_NE(_val1, _val2) ASSERT_TRUE((_val1) != (_val2))
+
+static struct vm_area_struct *alloc_vma(struct mm_struct *mm,
+                                       unsigned long start,
+                                       unsigned long end,
+                                       pgoff_t pgoff,
+                                       vm_flags_t flags)
+{
+       struct vm_area_struct *ret = vm_area_alloc(mm);
+
+       if (ret == NULL)
+               return NULL;
+
+       ret->vm_start = start;
+       ret->vm_end = end;
+       ret->vm_pgoff = pgoff;
+       ret->__vm_flags = flags;
+
+       return ret;
+}
+
+static bool test_simple_merge(void)
+{
+       struct vm_area_struct *vma;
+       unsigned long flags = VM_READ | VM_WRITE | VM_MAYREAD | VM_MAYWRITE;
+       struct mm_struct mm = {};
+       struct vm_area_struct *vma_left = alloc_vma(&mm, 0, 0x1000, 0, flags);
+       struct vm_area_struct *vma_middle = alloc_vma(&mm, 0x1000, 0x2000, 1, flags);
+       struct vm_area_struct *vma_right = alloc_vma(&mm, 0x2000, 0x3000, 2, flags);
+       VMA_ITERATOR(vmi, &mm, 0x1000);
+
+       ASSERT_FALSE(vma_link(&mm, vma_left));
+       ASSERT_FALSE(vma_link(&mm, vma_middle));
+       ASSERT_FALSE(vma_link(&mm, vma_right));
+
+       vma = vma_merge_new_vma(&vmi, vma_left, vma_middle, 0x1000,
+                               0x2000, 1);
+       ASSERT_NE(vma, NULL);
+
+       ASSERT_EQ(vma->vm_start, 0);
+       ASSERT_EQ(vma->vm_end, 0x3000);
+       ASSERT_EQ(vma->vm_pgoff, 0);
+       ASSERT_EQ(vma->vm_flags, flags);
+
+       vm_area_free(vma);
+       mtree_destroy(&mm.mm_mt);
+
+       return true;
+}
+
+static bool test_simple_modify(void)
+{
+       struct vm_area_struct *vma;
+       unsigned long flags = VM_READ | VM_WRITE | VM_MAYREAD | VM_MAYWRITE;
+       struct mm_struct mm = {};
+       struct vm_area_struct *init_vma = alloc_vma(&mm, 0, 0x3000, 0, flags);
+       VMA_ITERATOR(vmi, &mm, 0x1000);
+
+       ASSERT_FALSE(vma_link(&mm, init_vma));
+
+       /*
+        * The flags will not be changed, the vma_modify_flags() function
+        * performs the merge/split only.
+        */
+       vma = vma_modify_flags(&vmi, init_vma, init_vma,
+                              0x1000, 0x2000, VM_READ | VM_MAYREAD);
+       ASSERT_NE(vma, NULL);
+       /* We modify the provided VMA, and on split allocate new VMAs. */
+       ASSERT_EQ(vma, init_vma);
+
+       ASSERT_EQ(vma->vm_start, 0x1000);
+       ASSERT_EQ(vma->vm_end, 0x2000);
+       ASSERT_EQ(vma->vm_pgoff, 1);
+
+       /*
+        * Now walk through the three split VMAs and make sure they are as
+        * expected.
+        */
+
+       vma_iter_set(&vmi, 0);
+       vma = vma_iter_load(&vmi);
+
+       ASSERT_EQ(vma->vm_start, 0);
+       ASSERT_EQ(vma->vm_end, 0x1000);
+       ASSERT_EQ(vma->vm_pgoff, 0);
+
+       vm_area_free(vma);
+       vma_iter_clear(&vmi);
+
+       vma = vma_next(&vmi);
+
+       ASSERT_EQ(vma->vm_start, 0x1000);
+       ASSERT_EQ(vma->vm_end, 0x2000);
+       ASSERT_EQ(vma->vm_pgoff, 1);
+
+       vm_area_free(vma);
+       vma_iter_clear(&vmi);
+
+       vma = vma_next(&vmi);
+
+       ASSERT_EQ(vma->vm_start, 0x2000);
+       ASSERT_EQ(vma->vm_end, 0x3000);
+       ASSERT_EQ(vma->vm_pgoff, 2);
+
+       vm_area_free(vma);
+       mtree_destroy(&mm.mm_mt);
+
+       return true;
+}
+
+static bool test_simple_expand(void)
+{
+       unsigned long flags = VM_READ | VM_WRITE | VM_MAYREAD | VM_MAYWRITE;
+       struct mm_struct mm = {};
+       struct vm_area_struct *vma = alloc_vma(&mm, 0, 0x1000, 0, flags);
+       VMA_ITERATOR(vmi, &mm, 0);
+
+       ASSERT_FALSE(vma_link(&mm, vma));
+
+       ASSERT_FALSE(vma_expand(&vmi, vma, 0, 0x3000, 0, NULL));
+
+       ASSERT_EQ(vma->vm_start, 0);
+       ASSERT_EQ(vma->vm_end, 0x3000);
+       ASSERT_EQ(vma->vm_pgoff, 0);
+
+       vm_area_free(vma);
+       mtree_destroy(&mm.mm_mt);
+
+       return true;
+}
+
+static bool test_simple_shrink(void)
+{
+       unsigned long flags = VM_READ | VM_WRITE | VM_MAYREAD | VM_MAYWRITE;
+       struct mm_struct mm = {};
+       struct vm_area_struct *vma = alloc_vma(&mm, 0, 0x3000, 0, flags);
+       VMA_ITERATOR(vmi, &mm, 0);
+
+       ASSERT_FALSE(vma_link(&mm, vma));
+
+       ASSERT_FALSE(vma_shrink(&vmi, vma, 0, 0x1000, 0));
+
+       ASSERT_EQ(vma->vm_start, 0);
+       ASSERT_EQ(vma->vm_end, 0x1000);
+       ASSERT_EQ(vma->vm_pgoff, 0);
+
+       vm_area_free(vma);
+       mtree_destroy(&mm.mm_mt);
+
+       return true;
+}
+
+int main(void)
+{
+       int num_tests = 0, num_fail = 0;
+
+       maple_tree_init();
+
+#define TEST(name)                                                     \
+       do {                                                            \
+               num_tests++;                                            \
+               if (!test_##name()) {                                   \
+                       num_fail++;                                     \
+                       fprintf(stderr, "Test " #name " FAILED\n");     \
+               }                                                       \
+       } while (0)
+
+       TEST(simple_merge);
+       TEST(simple_modify);
+       TEST(simple_expand);
+       TEST(simple_shrink);
+
+#undef TEST
+
+       printf("%d tests run, %d passed, %d failed.\n",
+              num_tests, num_tests - num_fail, num_fail);
+
+       return num_fail == 0 ? EXIT_SUCCESS : EXIT_FAILURE;
+}
diff --git a/tools/testing/vma/vma_internal.h b/tools/testing/vma/vma_internal.h
new file mode 100644 (file)
index 0000000..093560e
--- /dev/null
@@ -0,0 +1,882 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * vma_internal.h
+ *
+ * Header providing userland wrappers and shims for the functionality provided
+ * by mm/vma_internal.h.
+ *
+ * We make the header guard the same as mm/vma_internal.h, so if this shim
+ * header is included, it precludes the inclusion of the kernel one.
+ */
+
+#ifndef __MM_VMA_INTERNAL_H
+#define __MM_VMA_INTERNAL_H
+
+#define __private
+#define __bitwise
+#define __randomize_layout
+
+#define CONFIG_MMU
+#define CONFIG_PER_VMA_LOCK
+
+#include <stdlib.h>
+
+#include <linux/list.h>
+#include <linux/maple_tree.h>
+#include <linux/mm.h>
+#include <linux/rbtree.h>
+#include <linux/rwsem.h>
+
+#define VM_WARN_ON(_expr) (WARN_ON(_expr))
+#define VM_WARN_ON_ONCE(_expr) (WARN_ON_ONCE(_expr))
+#define VM_BUG_ON(_expr) (BUG_ON(_expr))
+#define VM_BUG_ON_VMA(_expr, _vma) (BUG_ON(_expr))
+
+#define VM_NONE                0x00000000
+#define VM_READ                0x00000001
+#define VM_WRITE       0x00000002
+#define VM_EXEC                0x00000004
+#define VM_SHARED      0x00000008
+#define VM_MAYREAD     0x00000010
+#define VM_MAYWRITE    0x00000020
+#define VM_GROWSDOWN   0x00000100
+#define VM_PFNMAP      0x00000400
+#define VM_LOCKED      0x00002000
+#define VM_IO           0x00004000
+#define VM_DONTEXPAND  0x00040000
+#define VM_ACCOUNT     0x00100000
+#define VM_MIXEDMAP    0x10000000
+#define VM_STACK       VM_GROWSDOWN
+#define VM_SHADOW_STACK        VM_NONE
+#define VM_SOFTDIRTY   0
+
+#define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
+#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
+
+#define FIRST_USER_ADDRESS     0UL
+#define USER_PGTABLES_CEILING  0UL
+
+#define vma_policy(vma) NULL
+
+#define down_write_nest_lock(sem, nest_lock)
+
+#define pgprot_val(x)          ((x).pgprot)
+#define __pgprot(x)            ((pgprot_t) { (x) } )
+
+#define for_each_vma(__vmi, __vma)                                     \
+       while (((__vma) = vma_next(&(__vmi))) != NULL)
+
+/* The MM code likes to work with exclusive end addresses */
+#define for_each_vma_range(__vmi, __vma, __end)                                \
+       while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
+
+#define offset_in_page(p)      ((unsigned long)(p) & ~PAGE_MASK)
+
+#define PHYS_PFN(x)    ((unsigned long)((x) >> PAGE_SHIFT))
+
+#define test_and_set_bit(nr, addr) __test_and_set_bit(nr, addr)
+#define test_and_clear_bit(nr, addr) __test_and_clear_bit(nr, addr)
+
+#define TASK_SIZE ((1ul << 47)-PAGE_SIZE)
+
+#define AS_MM_ALL_LOCKS 2
+
+#define current NULL
+
+/* We hardcode this for now. */
+#define sysctl_max_map_count 0x1000000UL
+
+#define pgoff_t unsigned long
+typedef unsigned long  pgprotval_t;
+typedef struct pgprot { pgprotval_t pgprot; } pgprot_t;
+typedef unsigned long vm_flags_t;
+typedef __bitwise unsigned int vm_fault_t;
+
+typedef struct refcount_struct {
+       atomic_t refs;
+} refcount_t;
+
+struct kref {
+       refcount_t refcount;
+};
+
+struct anon_vma {
+       struct anon_vma *root;
+       struct rb_root_cached rb_root;
+};
+
+struct anon_vma_chain {
+       struct anon_vma *anon_vma;
+       struct list_head same_vma;
+};
+
+struct anon_vma_name {
+       struct kref kref;
+       /* The name needs to be at the end because it is dynamically sized. */
+       char name[];
+};
+
+struct vma_iterator {
+       struct ma_state mas;
+};
+
+#define VMA_ITERATOR(name, __mm, __addr)                               \
+       struct vma_iterator name = {                                    \
+               .mas = {                                                \
+                       .tree = &(__mm)->mm_mt,                         \
+                       .index = __addr,                                \
+                       .node = NULL,                                   \
+                       .status = ma_start,                             \
+               },                                                      \
+       }
+
+struct address_space {
+       struct rb_root_cached   i_mmap;
+       unsigned long           flags;
+       atomic_t                i_mmap_writable;
+};
+
+struct vm_userfaultfd_ctx {};
+struct mempolicy {};
+struct mmu_gather {};
+struct mutex {};
+#define DEFINE_MUTEX(mutexname) \
+       struct mutex mutexname = {}
+
+struct mm_struct {
+       struct maple_tree mm_mt;
+       int map_count;                  /* number of VMAs */
+       unsigned long total_vm;    /* Total pages mapped */
+       unsigned long locked_vm;   /* Pages that have PG_mlocked set */
+       unsigned long data_vm;     /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
+       unsigned long exec_vm;     /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
+       unsigned long stack_vm;    /* VM_STACK */
+};
+
+struct vma_lock {
+       struct rw_semaphore lock;
+};
+
+
+struct file {
+       struct address_space    *f_mapping;
+};
+
+struct vm_area_struct {
+       /* The first cache line has the info for VMA tree walking. */
+
+       union {
+               struct {
+                       /* VMA covers [vm_start; vm_end) addresses within mm */
+                       unsigned long vm_start;
+                       unsigned long vm_end;
+               };
+#ifdef CONFIG_PER_VMA_LOCK
+               struct rcu_head vm_rcu; /* Used for deferred freeing. */
+#endif
+       };
+
+       struct mm_struct *vm_mm;        /* The address space we belong to. */
+       pgprot_t vm_page_prot;          /* Access permissions of this VMA. */
+
+       /*
+        * Flags, see mm.h.
+        * To modify use vm_flags_{init|reset|set|clear|mod} functions.
+        */
+       union {
+               const vm_flags_t vm_flags;
+               vm_flags_t __private __vm_flags;
+       };
+
+#ifdef CONFIG_PER_VMA_LOCK
+       /* Flag to indicate areas detached from the mm->mm_mt tree */
+       bool detached;
+
+       /*
+        * Can only be written (using WRITE_ONCE()) while holding both:
+        *  - mmap_lock (in write mode)
+        *  - vm_lock->lock (in write mode)
+        * Can be read reliably while holding one of:
+        *  - mmap_lock (in read or write mode)
+        *  - vm_lock->lock (in read or write mode)
+        * Can be read unreliably (using READ_ONCE()) for pessimistic bailout
+        * while holding nothing (except RCU to keep the VMA struct allocated).
+        *
+        * This sequence counter is explicitly allowed to overflow; sequence
+        * counter reuse can only lead to occasional unnecessary use of the
+        * slowpath.
+        */
+       int vm_lock_seq;
+       struct vma_lock *vm_lock;
+#endif
+
+       /*
+        * For areas with an address space and backing store,
+        * linkage into the address_space->i_mmap interval tree.
+        *
+        */
+       struct {
+               struct rb_node rb;
+               unsigned long rb_subtree_last;
+       } shared;
+
+       /*
+        * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
+        * list, after a COW of one of the file pages.  A MAP_SHARED vma
+        * can only be in the i_mmap tree.  An anonymous MAP_PRIVATE, stack
+        * or brk vma (with NULL file) can only be in an anon_vma list.
+        */
+       struct list_head anon_vma_chain; /* Serialized by mmap_lock &
+                                         * page_table_lock */
+       struct anon_vma *anon_vma;      /* Serialized by page_table_lock */
+
+       /* Function pointers to deal with this struct. */
+       const struct vm_operations_struct *vm_ops;
+
+       /* Information about our backing store: */
+       unsigned long vm_pgoff;         /* Offset (within vm_file) in PAGE_SIZE
+                                          units */
+       struct file * vm_file;          /* File we map to (can be NULL). */
+       void * vm_private_data;         /* was vm_pte (shared mem) */
+
+#ifdef CONFIG_ANON_VMA_NAME
+       /*
+        * For private and shared anonymous mappings, a pointer to a null
+        * terminated string containing the name given to the vma, or NULL if
+        * unnamed. Serialized by mmap_lock. Use anon_vma_name to access.
+        */
+       struct anon_vma_name *anon_name;
+#endif
+#ifdef CONFIG_SWAP
+       atomic_long_t swap_readahead_info;
+#endif
+#ifndef CONFIG_MMU
+       struct vm_region *vm_region;    /* NOMMU mapping region */
+#endif
+#ifdef CONFIG_NUMA
+       struct mempolicy *vm_policy;    /* NUMA policy for the VMA */
+#endif
+#ifdef CONFIG_NUMA_BALANCING
+       struct vma_numab_state *numab_state;    /* NUMA Balancing state */
+#endif
+       struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
+} __randomize_layout;
+
+struct vm_fault {};
+
+struct vm_operations_struct {
+       void (*open)(struct vm_area_struct * area);
+       /**
+        * @close: Called when the VMA is being removed from the MM.
+        * Context: User context.  May sleep.  Caller holds mmap_lock.
+        */
+       void (*close)(struct vm_area_struct * area);
+       /* Called any time before splitting to check if it's allowed */
+       int (*may_split)(struct vm_area_struct *area, unsigned long addr);
+       int (*mremap)(struct vm_area_struct *area);
+       /*
+        * Called by mprotect() to make driver-specific permission
+        * checks before mprotect() is finalised.   The VMA must not
+        * be modified.  Returns 0 if mprotect() can proceed.
+        */
+       int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
+                       unsigned long end, unsigned long newflags);
+       vm_fault_t (*fault)(struct vm_fault *vmf);
+       vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
+       vm_fault_t (*map_pages)(struct vm_fault *vmf,
+                       pgoff_t start_pgoff, pgoff_t end_pgoff);
+       unsigned long (*pagesize)(struct vm_area_struct * area);
+
+       /* notification that a previously read-only page is about to become
+        * writable, if an error is returned it will cause a SIGBUS */
+       vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
+
+       /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
+       vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
+
+       /* called by access_process_vm when get_user_pages() fails, typically
+        * for use by special VMAs. See also generic_access_phys() for a generic
+        * implementation useful for any iomem mapping.
+        */
+       int (*access)(struct vm_area_struct *vma, unsigned long addr,
+                     void *buf, int len, int write);
+
+       /* Called by the /proc/PID/maps code to ask the vma whether it
+        * has a special name.  Returning non-NULL will also cause this
+        * vma to be dumped unconditionally. */
+       const char *(*name)(struct vm_area_struct *vma);
+
+#ifdef CONFIG_NUMA
+       /*
+        * set_policy() op must add a reference to any non-NULL @new mempolicy
+        * to hold the policy upon return.  Caller should pass NULL @new to
+        * remove a policy and fall back to surrounding context--i.e. do not
+        * install a MPOL_DEFAULT policy, nor the task or system default
+        * mempolicy.
+        */
+       int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
+
+       /*
+        * get_policy() op must add reference [mpol_get()] to any policy at
+        * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
+        * in mm/mempolicy.c will do this automatically.
+        * get_policy() must NOT add a ref if the policy at (vma,addr) is not
+        * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
+        * If no [shared/vma] mempolicy exists at the addr, get_policy() op
+        * must return NULL--i.e., do not "fallback" to task or system default
+        * policy.
+        */
+       struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
+                                       unsigned long addr, pgoff_t *ilx);
+#endif
+       /*
+        * Called by vm_normal_page() for special PTEs to find the
+        * page for @addr.  This is useful if the default behavior
+        * (using pte_page()) would not find the correct page.
+        */
+       struct page *(*find_special_page)(struct vm_area_struct *vma,
+                                         unsigned long addr);
+};
+
+static inline void vma_iter_invalidate(struct vma_iterator *vmi)
+{
+       mas_pause(&vmi->mas);
+}
+
+static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
+{
+       return __pgprot(pgprot_val(oldprot) | pgprot_val(newprot));
+}
+
+static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
+{
+       return __pgprot(vm_flags);
+}
+
+static inline bool is_shared_maywrite(vm_flags_t vm_flags)
+{
+       return (vm_flags & (VM_SHARED | VM_MAYWRITE)) ==
+               (VM_SHARED | VM_MAYWRITE);
+}
+
+static inline bool vma_is_shared_maywrite(struct vm_area_struct *vma)
+{
+       return is_shared_maywrite(vma->vm_flags);
+}
+
+static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
+{
+       /*
+        * Uses mas_find() to get the first VMA when the iterator starts.
+        * Calling mas_next() could skip the first entry.
+        */
+       return mas_find(&vmi->mas, ULONG_MAX);
+}
+
+static inline bool vma_lock_alloc(struct vm_area_struct *vma)
+{
+       vma->vm_lock = calloc(1, sizeof(struct vma_lock));
+
+       if (!vma->vm_lock)
+               return false;
+
+       init_rwsem(&vma->vm_lock->lock);
+       vma->vm_lock_seq = -1;
+
+       return true;
+}
+
+static inline void vma_assert_write_locked(struct vm_area_struct *);
+static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
+{
+       /* When detaching vma should be write-locked */
+       if (detached)
+               vma_assert_write_locked(vma);
+       vma->detached = detached;
+}
+
+extern const struct vm_operations_struct vma_dummy_vm_ops;
+
+static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
+{
+       memset(vma, 0, sizeof(*vma));
+       vma->vm_mm = mm;
+       vma->vm_ops = &vma_dummy_vm_ops;
+       INIT_LIST_HEAD(&vma->anon_vma_chain);
+       vma_mark_detached(vma, false);
+}
+
+static inline struct vm_area_struct *vm_area_alloc(struct mm_struct *mm)
+{
+       struct vm_area_struct *vma = calloc(1, sizeof(struct vm_area_struct));
+
+       if (!vma)
+               return NULL;
+
+       vma_init(vma, mm);
+       if (!vma_lock_alloc(vma)) {
+               free(vma);
+               return NULL;
+       }
+
+       return vma;
+}
+
+static inline struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
+{
+       struct vm_area_struct *new = calloc(1, sizeof(struct vm_area_struct));
+
+       if (!new)
+               return NULL;
+
+       memcpy(new, orig, sizeof(*new));
+       if (!vma_lock_alloc(new)) {
+               free(new);
+               return NULL;
+       }
+       INIT_LIST_HEAD(&new->anon_vma_chain);
+
+       return new;
+}
+
+/*
+ * These are defined in vma.h, but sadly vm_stat_account() is referenced by
+ * kernel/fork.c, so we have to these broadly available there, and temporarily
+ * define them here to resolve the dependency cycle.
+ */
+
+#define is_exec_mapping(flags) \
+       ((flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC)
+
+#define is_stack_mapping(flags) \
+       (((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK))
+
+#define is_data_mapping(flags) \
+       ((flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE)
+
+static inline void vm_stat_account(struct mm_struct *mm, vm_flags_t flags,
+                                  long npages)
+{
+       WRITE_ONCE(mm->total_vm, READ_ONCE(mm->total_vm)+npages);
+
+       if (is_exec_mapping(flags))
+               mm->exec_vm += npages;
+       else if (is_stack_mapping(flags))
+               mm->stack_vm += npages;
+       else if (is_data_mapping(flags))
+               mm->data_vm += npages;
+}
+
+#undef is_exec_mapping
+#undef is_stack_mapping
+#undef is_data_mapping
+
+/* Currently stubbed but we may later wish to un-stub. */
+static inline void vm_acct_memory(long pages);
+static inline void vm_unacct_memory(long pages)
+{
+       vm_acct_memory(-pages);
+}
+
+static inline void mapping_allow_writable(struct address_space *mapping)
+{
+       atomic_inc(&mapping->i_mmap_writable);
+}
+
+static inline void vma_set_range(struct vm_area_struct *vma,
+                                unsigned long start, unsigned long end,
+                                pgoff_t pgoff)
+{
+       vma->vm_start = start;
+       vma->vm_end = end;
+       vma->vm_pgoff = pgoff;
+}
+
+static inline
+struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
+{
+       return mas_find(&vmi->mas, max - 1);
+}
+
+static inline int vma_iter_clear_gfp(struct vma_iterator *vmi,
+                       unsigned long start, unsigned long end, gfp_t gfp)
+{
+       __mas_set_range(&vmi->mas, start, end - 1);
+       mas_store_gfp(&vmi->mas, NULL, gfp);
+       if (unlikely(mas_is_err(&vmi->mas)))
+               return -ENOMEM;
+
+       return 0;
+}
+
+static inline void mmap_assert_locked(struct mm_struct *);
+static inline struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
+                                               unsigned long start_addr,
+                                               unsigned long end_addr)
+{
+       unsigned long index = start_addr;
+
+       mmap_assert_locked(mm);
+       return mt_find(&mm->mm_mt, &index, end_addr - 1);
+}
+
+static inline
+struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
+{
+       return mtree_load(&mm->mm_mt, addr);
+}
+
+static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
+{
+       return mas_prev(&vmi->mas, 0);
+}
+
+static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
+{
+       mas_set(&vmi->mas, addr);
+}
+
+static inline bool vma_is_anonymous(struct vm_area_struct *vma)
+{
+       return !vma->vm_ops;
+}
+
+/* Defined in vma.h, so temporarily define here to avoid circular dependency. */
+#define vma_iter_load(vmi) \
+       mas_walk(&(vmi)->mas)
+
+static inline struct vm_area_struct *
+find_vma_prev(struct mm_struct *mm, unsigned long addr,
+                       struct vm_area_struct **pprev)
+{
+       struct vm_area_struct *vma;
+       VMA_ITERATOR(vmi, mm, addr);
+
+       vma = vma_iter_load(&vmi);
+       *pprev = vma_prev(&vmi);
+       if (!vma)
+               vma = vma_next(&vmi);
+       return vma;
+}
+
+#undef vma_iter_load
+
+static inline void vma_iter_init(struct vma_iterator *vmi,
+               struct mm_struct *mm, unsigned long addr)
+{
+       mas_init(&vmi->mas, &mm->mm_mt, addr);
+}
+
+/* Stubbed functions. */
+
+static inline struct anon_vma_name *anon_vma_name(struct vm_area_struct *vma)
+{
+       return NULL;
+}
+
+static inline bool is_mergeable_vm_userfaultfd_ctx(struct vm_area_struct *vma,
+                                       struct vm_userfaultfd_ctx vm_ctx)
+{
+       return true;
+}
+
+static inline bool anon_vma_name_eq(struct anon_vma_name *anon_name1,
+                                   struct anon_vma_name *anon_name2)
+{
+       return true;
+}
+
+static inline void might_sleep(void)
+{
+}
+
+static inline unsigned long vma_pages(struct vm_area_struct *vma)
+{
+       return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
+}
+
+static inline void fput(struct file *)
+{
+}
+
+static inline void mpol_put(struct mempolicy *)
+{
+}
+
+static inline void vma_lock_free(struct vm_area_struct *vma)
+{
+       free(vma->vm_lock);
+}
+
+static inline void __vm_area_free(struct vm_area_struct *vma)
+{
+       vma_lock_free(vma);
+       free(vma);
+}
+
+static inline void vm_area_free(struct vm_area_struct *vma)
+{
+       __vm_area_free(vma);
+}
+
+static inline void lru_add_drain(void)
+{
+}
+
+static inline void tlb_gather_mmu(struct mmu_gather *, struct mm_struct *)
+{
+}
+
+static inline void update_hiwater_rss(struct mm_struct *)
+{
+}
+
+static inline void update_hiwater_vm(struct mm_struct *)
+{
+}
+
+static inline void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
+                     struct vm_area_struct *vma, unsigned long start_addr,
+                     unsigned long end_addr, unsigned long tree_end,
+                     bool mm_wr_locked)
+{
+       (void)tlb;
+       (void)mas;
+       (void)vma;
+       (void)start_addr;
+       (void)end_addr;
+       (void)tree_end;
+       (void)mm_wr_locked;
+}
+
+static inline void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
+                  struct vm_area_struct *vma, unsigned long floor,
+                  unsigned long ceiling, bool mm_wr_locked)
+{
+       (void)tlb;
+       (void)mas;
+       (void)vma;
+       (void)floor;
+       (void)ceiling;
+       (void)mm_wr_locked;
+}
+
+static inline void mapping_unmap_writable(struct address_space *)
+{
+}
+
+static inline void flush_dcache_mmap_lock(struct address_space *)
+{
+}
+
+static inline void tlb_finish_mmu(struct mmu_gather *)
+{
+}
+
+static inline void get_file(struct file *)
+{
+}
+
+static inline int vma_dup_policy(struct vm_area_struct *, struct vm_area_struct *)
+{
+       return 0;
+}
+
+static inline int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *)
+{
+       return 0;
+}
+
+static inline void vma_start_write(struct vm_area_struct *)
+{
+}
+
+static inline void vma_adjust_trans_huge(struct vm_area_struct *vma,
+                                        unsigned long start,
+                                        unsigned long end,
+                                        long adjust_next)
+{
+       (void)vma;
+       (void)start;
+       (void)end;
+       (void)adjust_next;
+}
+
+static inline void vma_iter_free(struct vma_iterator *vmi)
+{
+       mas_destroy(&vmi->mas);
+}
+
+static inline void vm_acct_memory(long pages)
+{
+}
+
+static inline void vma_interval_tree_insert(struct vm_area_struct *,
+                                           struct rb_root_cached *)
+{
+}
+
+static inline void vma_interval_tree_remove(struct vm_area_struct *,
+                                           struct rb_root_cached *)
+{
+}
+
+static inline void flush_dcache_mmap_unlock(struct address_space *)
+{
+}
+
+static inline void anon_vma_interval_tree_insert(struct anon_vma_chain*,
+                                                struct rb_root_cached *)
+{
+}
+
+static inline void anon_vma_interval_tree_remove(struct anon_vma_chain*,
+                                                struct rb_root_cached *)
+{
+}
+
+static inline void uprobe_mmap(struct vm_area_struct *)
+{
+}
+
+static inline void uprobe_munmap(struct vm_area_struct *vma,
+                                unsigned long start, unsigned long end)
+{
+       (void)vma;
+       (void)start;
+       (void)end;
+}
+
+static inline void i_mmap_lock_write(struct address_space *)
+{
+}
+
+static inline void anon_vma_lock_write(struct anon_vma *)
+{
+}
+
+static inline void vma_assert_write_locked(struct vm_area_struct *)
+{
+}
+
+static inline void unlink_anon_vmas(struct vm_area_struct *)
+{
+}
+
+static inline void anon_vma_unlock_write(struct anon_vma *)
+{
+}
+
+static inline void i_mmap_unlock_write(struct address_space *)
+{
+}
+
+static inline void anon_vma_merge(struct vm_area_struct *,
+                                 struct vm_area_struct *)
+{
+}
+
+static inline int userfaultfd_unmap_prep(struct vm_area_struct *vma,
+                                        unsigned long start,
+                                        unsigned long end,
+                                        struct list_head *unmaps)
+{
+       (void)vma;
+       (void)start;
+       (void)end;
+       (void)unmaps;
+
+       return 0;
+}
+
+static inline void mmap_write_downgrade(struct mm_struct *)
+{
+}
+
+static inline void mmap_read_unlock(struct mm_struct *)
+{
+}
+
+static inline void mmap_write_unlock(struct mm_struct *)
+{
+}
+
+static inline bool can_modify_mm(struct mm_struct *mm,
+                                unsigned long start,
+                                unsigned long end)
+{
+       (void)mm;
+       (void)start;
+       (void)end;
+
+       return true;
+}
+
+static inline void arch_unmap(struct mm_struct *mm,
+                                unsigned long start,
+                                unsigned long end)
+{
+       (void)mm;
+       (void)start;
+       (void)end;
+}
+
+static inline void mmap_assert_locked(struct mm_struct *)
+{
+}
+
+static inline bool mpol_equal(struct mempolicy *, struct mempolicy *)
+{
+       return true;
+}
+
+static inline void khugepaged_enter_vma(struct vm_area_struct *vma,
+                         unsigned long vm_flags)
+{
+       (void)vma;
+       (void)vm_flags;
+}
+
+static inline bool mapping_can_writeback(struct address_space *)
+{
+       return true;
+}
+
+static inline bool is_vm_hugetlb_page(struct vm_area_struct *)
+{
+       return false;
+}
+
+static inline bool vma_soft_dirty_enabled(struct vm_area_struct *)
+{
+       return false;
+}
+
+static inline bool userfaultfd_wp(struct vm_area_struct *)
+{
+       return false;
+}
+
+static inline void mmap_assert_write_locked(struct mm_struct *)
+{
+}
+
+static inline void mutex_lock(struct mutex *)
+{
+}
+
+static inline void mutex_unlock(struct mutex *)
+{
+}
+
+static inline bool mutex_is_locked(struct mutex *)
+{
+       return true;
+}
+
+static inline bool signal_pending(void *)
+{
+       return false;
+}
+
+#endif /* __MM_VMA_INTERNAL_H */