mm: restructure memfd code
authorMike Kravetz <mike.kravetz@oracle.com>
Fri, 8 Jun 2018 00:06:01 +0000 (17:06 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 8 Jun 2018 00:34:35 +0000 (17:34 -0700)
With the addition of memfd hugetlbfs support, we now have the situation
where memfd depends on TMPFS -or- HUGETLBFS.  Previously, memfd was only
supported on tmpfs, so it made sense that the code resided in shmem.c.
In the current code, memfd is only functional if TMPFS is defined.  If
HUGETLFS is defined and TMPFS is not defined, then memfd functionality
will not be available for hugetlbfs.  This does not cause BUGs, just a
lack of potentially desired functionality.

Code is restructured in the following way:
- include/linux/memfd.h is a new file containing memfd specific
  definitions previously contained in shmem_fs.h.
- mm/memfd.c is a new file containing memfd specific code previously
  contained in shmem.c.
- memfd specific code is removed from shmem_fs.h and shmem.c.
- A new config option MEMFD_CREATE is added that is defined if TMPFS
  or HUGETLBFS is defined.

No functional changes are made to the code: restructuring only.

Link: http://lkml.kernel.org/r/20180415182119.4517-4-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Khalid Aziz <khalid.aziz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marc-Andr Lureau <marcandre.lureau@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/Kconfig
fs/fcntl.c
include/linux/memfd.h [new file with mode: 0644]
include/linux/shmem_fs.h
mm/Makefile
mm/memfd.c [new file with mode: 0644]
mm/shmem.c

index ac4ac90..51f78a2 100644 (file)
@@ -203,6 +203,9 @@ config HUGETLBFS
 config HUGETLB_PAGE
        def_bool HUGETLBFS
 
+config MEMFD_CREATE
+       def_bool TMPFS || HUGETLBFS
+
 config ARCH_HAS_GIGANTIC_PAGE
        bool
 
index c421694..12273b6 100644 (file)
@@ -23,7 +23,7 @@
 #include <linux/rcupdate.h>
 #include <linux/pid_namespace.h>
 #include <linux/user_namespace.h>
-#include <linux/shmem_fs.h>
+#include <linux/memfd.h>
 #include <linux/compat.h>
 
 #include <linux/poll.h>
diff --git a/include/linux/memfd.h b/include/linux/memfd.h
new file mode 100644 (file)
index 0000000..4f16004
--- /dev/null
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __LINUX_MEMFD_H
+#define __LINUX_MEMFD_H
+
+#include <linux/file.h>
+
+#ifdef CONFIG_MEMFD_CREATE
+extern long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg);
+#else
+static inline long memfd_fcntl(struct file *f, unsigned int c, unsigned long a)
+{
+       return -EINVAL;
+}
+#endif
+
+#endif /* __LINUX_MEMFD_H */
index 73b5e65..f155dc6 100644 (file)
@@ -110,19 +110,6 @@ static inline bool shmem_file(struct file *file)
 extern bool shmem_charge(struct inode *inode, long pages);
 extern void shmem_uncharge(struct inode *inode, long pages);
 
-#ifdef CONFIG_TMPFS
-
-extern long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg);
-
-#else
-
-static inline long memfd_fcntl(struct file *f, unsigned int c, unsigned long a)
-{
-       return -EINVAL;
-}
-
-#endif
-
 #ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
 extern bool shmem_huge_enabled(struct vm_area_struct *vma);
 #else
index b4e54a9..8716bda 100644 (file)
@@ -105,3 +105,4 @@ obj-$(CONFIG_DEBUG_PAGE_REF) += debug_page_ref.o
 obj-$(CONFIG_HARDENED_USERCOPY) += usercopy.o
 obj-$(CONFIG_PERCPU_STATS) += percpu-stats.o
 obj-$(CONFIG_HMM) += hmm.o
+obj-$(CONFIG_MEMFD_CREATE) += memfd.o
diff --git a/mm/memfd.c b/mm/memfd.c
new file mode 100644 (file)
index 0000000..2706951
--- /dev/null
@@ -0,0 +1,345 @@
+/*
+ * memfd_create system call and file sealing support
+ *
+ * Code was originally included in shmem.c, and broken out to facilitate
+ * use by hugetlbfs as well as tmpfs.
+ *
+ * This file is released under the GPL.
+ */
+
+#include <linux/fs.h>
+#include <linux/vfs.h>
+#include <linux/pagemap.h>
+#include <linux/file.h>
+#include <linux/mm.h>
+#include <linux/sched/signal.h>
+#include <linux/khugepaged.h>
+#include <linux/syscalls.h>
+#include <linux/hugetlb.h>
+#include <linux/shmem_fs.h>
+#include <linux/memfd.h>
+#include <uapi/linux/memfd.h>
+
+/*
+ * We need a tag: a new tag would expand every radix_tree_node by 8 bytes,
+ * so reuse a tag which we firmly believe is never set or cleared on tmpfs
+ * or hugetlbfs because they are memory only filesystems.
+ */
+#define MEMFD_TAG_PINNED        PAGECACHE_TAG_TOWRITE
+#define LAST_SCAN               4       /* about 150ms max */
+
+static void memfd_tag_pins(struct address_space *mapping)
+{
+       struct radix_tree_iter iter;
+       void __rcu **slot;
+       pgoff_t start;
+       struct page *page;
+
+       lru_add_drain();
+       start = 0;
+       rcu_read_lock();
+
+       radix_tree_for_each_slot(slot, &mapping->i_pages, &iter, start) {
+               page = radix_tree_deref_slot(slot);
+               if (!page || radix_tree_exception(page)) {
+                       if (radix_tree_deref_retry(page)) {
+                               slot = radix_tree_iter_retry(&iter);
+                               continue;
+                       }
+               } else if (page_count(page) - page_mapcount(page) > 1) {
+                       xa_lock_irq(&mapping->i_pages);
+                       radix_tree_tag_set(&mapping->i_pages, iter.index,
+                                          MEMFD_TAG_PINNED);
+                       xa_unlock_irq(&mapping->i_pages);
+               }
+
+               if (need_resched()) {
+                       slot = radix_tree_iter_resume(slot, &iter);
+                       cond_resched_rcu();
+               }
+       }
+       rcu_read_unlock();
+}
+
+/*
+ * Setting SEAL_WRITE requires us to verify there's no pending writer. However,
+ * via get_user_pages(), drivers might have some pending I/O without any active
+ * user-space mappings (eg., direct-IO, AIO). Therefore, we look at all pages
+ * and see whether it has an elevated ref-count. If so, we tag them and wait for
+ * them to be dropped.
+ * The caller must guarantee that no new user will acquire writable references
+ * to those pages to avoid races.
+ */
+static int memfd_wait_for_pins(struct address_space *mapping)
+{
+       struct radix_tree_iter iter;
+       void __rcu **slot;
+       pgoff_t start;
+       struct page *page;
+       int error, scan;
+
+       memfd_tag_pins(mapping);
+
+       error = 0;
+       for (scan = 0; scan <= LAST_SCAN; scan++) {
+               if (!radix_tree_tagged(&mapping->i_pages, MEMFD_TAG_PINNED))
+                       break;
+
+               if (!scan)
+                       lru_add_drain_all();
+               else if (schedule_timeout_killable((HZ << scan) / 200))
+                       scan = LAST_SCAN;
+
+               start = 0;
+               rcu_read_lock();
+               radix_tree_for_each_tagged(slot, &mapping->i_pages, &iter,
+                                          start, MEMFD_TAG_PINNED) {
+
+                       page = radix_tree_deref_slot(slot);
+                       if (radix_tree_exception(page)) {
+                               if (radix_tree_deref_retry(page)) {
+                                       slot = radix_tree_iter_retry(&iter);
+                                       continue;
+                               }
+
+                               page = NULL;
+                       }
+
+                       if (page &&
+                           page_count(page) - page_mapcount(page) != 1) {
+                               if (scan < LAST_SCAN)
+                                       goto continue_resched;
+
+                               /*
+                                * On the last scan, we clean up all those tags
+                                * we inserted; but make a note that we still
+                                * found pages pinned.
+                                */
+                               error = -EBUSY;
+                       }
+
+                       xa_lock_irq(&mapping->i_pages);
+                       radix_tree_tag_clear(&mapping->i_pages,
+                                            iter.index, MEMFD_TAG_PINNED);
+                       xa_unlock_irq(&mapping->i_pages);
+continue_resched:
+                       if (need_resched()) {
+                               slot = radix_tree_iter_resume(slot, &iter);
+                               cond_resched_rcu();
+                       }
+               }
+               rcu_read_unlock();
+       }
+
+       return error;
+}
+
+static unsigned int *memfd_file_seals_ptr(struct file *file)
+{
+       if (shmem_file(file))
+               return &SHMEM_I(file_inode(file))->seals;
+
+#ifdef CONFIG_HUGETLBFS
+       if (is_file_hugepages(file))
+               return &HUGETLBFS_I(file_inode(file))->seals;
+#endif
+
+       return NULL;
+}
+
+#define F_ALL_SEALS (F_SEAL_SEAL | \
+                    F_SEAL_SHRINK | \
+                    F_SEAL_GROW | \
+                    F_SEAL_WRITE)
+
+static int memfd_add_seals(struct file *file, unsigned int seals)
+{
+       struct inode *inode = file_inode(file);
+       unsigned int *file_seals;
+       int error;
+
+       /*
+        * SEALING
+        * Sealing allows multiple parties to share a tmpfs or hugetlbfs file
+        * but restrict access to a specific subset of file operations. Seals
+        * can only be added, but never removed. This way, mutually untrusted
+        * parties can share common memory regions with a well-defined policy.
+        * A malicious peer can thus never perform unwanted operations on a
+        * shared object.
+        *
+        * Seals are only supported on special tmpfs or hugetlbfs files and
+        * always affect the whole underlying inode. Once a seal is set, it
+        * may prevent some kinds of access to the file. Currently, the
+        * following seals are defined:
+        *   SEAL_SEAL: Prevent further seals from being set on this file
+        *   SEAL_SHRINK: Prevent the file from shrinking
+        *   SEAL_GROW: Prevent the file from growing
+        *   SEAL_WRITE: Prevent write access to the file
+        *
+        * As we don't require any trust relationship between two parties, we
+        * must prevent seals from being removed. Therefore, sealing a file
+        * only adds a given set of seals to the file, it never touches
+        * existing seals. Furthermore, the "setting seals"-operation can be
+        * sealed itself, which basically prevents any further seal from being
+        * added.
+        *
+        * Semantics of sealing are only defined on volatile files. Only
+        * anonymous tmpfs and hugetlbfs files support sealing. More
+        * importantly, seals are never written to disk. Therefore, there's
+        * no plan to support it on other file types.
+        */
+
+       if (!(file->f_mode & FMODE_WRITE))
+               return -EPERM;
+       if (seals & ~(unsigned int)F_ALL_SEALS)
+               return -EINVAL;
+
+       inode_lock(inode);
+
+       file_seals = memfd_file_seals_ptr(file);
+       if (!file_seals) {
+               error = -EINVAL;
+               goto unlock;
+       }
+
+       if (*file_seals & F_SEAL_SEAL) {
+               error = -EPERM;
+               goto unlock;
+       }
+
+       if ((seals & F_SEAL_WRITE) && !(*file_seals & F_SEAL_WRITE)) {
+               error = mapping_deny_writable(file->f_mapping);
+               if (error)
+                       goto unlock;
+
+               error = memfd_wait_for_pins(file->f_mapping);
+               if (error) {
+                       mapping_allow_writable(file->f_mapping);
+                       goto unlock;
+               }
+       }
+
+       *file_seals |= seals;
+       error = 0;
+
+unlock:
+       inode_unlock(inode);
+       return error;
+}
+
+static int memfd_get_seals(struct file *file)
+{
+       unsigned int *seals = memfd_file_seals_ptr(file);
+
+       return seals ? *seals : -EINVAL;
+}
+
+long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+       long error;
+
+       switch (cmd) {
+       case F_ADD_SEALS:
+               /* disallow upper 32bit */
+               if (arg > UINT_MAX)
+                       return -EINVAL;
+
+               error = memfd_add_seals(file, arg);
+               break;
+       case F_GET_SEALS:
+               error = memfd_get_seals(file);
+               break;
+       default:
+               error = -EINVAL;
+               break;
+       }
+
+       return error;
+}
+
+#define MFD_NAME_PREFIX "memfd:"
+#define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
+#define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
+
+#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB)
+
+SYSCALL_DEFINE2(memfd_create,
+               const char __user *, uname,
+               unsigned int, flags)
+{
+       unsigned int *file_seals;
+       struct file *file;
+       int fd, error;
+       char *name;
+       long len;
+
+       if (!(flags & MFD_HUGETLB)) {
+               if (flags & ~(unsigned int)MFD_ALL_FLAGS)
+                       return -EINVAL;
+       } else {
+               /* Allow huge page size encoding in flags. */
+               if (flags & ~(unsigned int)(MFD_ALL_FLAGS |
+                               (MFD_HUGE_MASK << MFD_HUGE_SHIFT)))
+                       return -EINVAL;
+       }
+
+       /* length includes terminating zero */
+       len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
+       if (len <= 0)
+               return -EFAULT;
+       if (len > MFD_NAME_MAX_LEN + 1)
+               return -EINVAL;
+
+       name = kmalloc(len + MFD_NAME_PREFIX_LEN, GFP_KERNEL);
+       if (!name)
+               return -ENOMEM;
+
+       strcpy(name, MFD_NAME_PREFIX);
+       if (copy_from_user(&name[MFD_NAME_PREFIX_LEN], uname, len)) {
+               error = -EFAULT;
+               goto err_name;
+       }
+
+       /* terminating-zero may have changed after strnlen_user() returned */
+       if (name[len + MFD_NAME_PREFIX_LEN - 1]) {
+               error = -EFAULT;
+               goto err_name;
+       }
+
+       fd = get_unused_fd_flags((flags & MFD_CLOEXEC) ? O_CLOEXEC : 0);
+       if (fd < 0) {
+               error = fd;
+               goto err_name;
+       }
+
+       if (flags & MFD_HUGETLB) {
+               struct user_struct *user = NULL;
+
+               file = hugetlb_file_setup(name, 0, VM_NORESERVE, &user,
+                                       HUGETLB_ANONHUGE_INODE,
+                                       (flags >> MFD_HUGE_SHIFT) &
+                                       MFD_HUGE_MASK);
+       } else
+               file = shmem_file_setup(name, 0, VM_NORESERVE);
+       if (IS_ERR(file)) {
+               error = PTR_ERR(file);
+               goto err_fd;
+       }
+       file->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
+       file->f_flags |= O_RDWR | O_LARGEFILE;
+
+       if (flags & MFD_ALLOW_SEALING) {
+               file_seals = memfd_file_seals_ptr(file);
+               *file_seals &= ~F_SEAL_SEAL;
+       }
+
+       fd_install(fd, file);
+       kfree(name);
+       return fd;
+
+err_fd:
+       put_unused_fd(fd);
+err_name:
+       kfree(name);
+       return error;
+}
index ceba031..cb4d7fa 100644 (file)
@@ -2618,243 +2618,6 @@ static loff_t shmem_file_llseek(struct file *file, loff_t offset, int whence)
        return offset;
 }
 
-/*
- * We need a tag: a new tag would expand every radix_tree_node by 8 bytes,
- * so reuse a tag which we firmly believe is never set or cleared on tmpfs
- * or hugetlbfs because they are memory only filesystems.
- */
-#define MEMFD_TAG_PINNED        PAGECACHE_TAG_TOWRITE
-#define LAST_SCAN               4       /* about 150ms max */
-
-static void memfd_tag_pins(struct address_space *mapping)
-{
-       struct radix_tree_iter iter;
-       void __rcu **slot;
-       pgoff_t start;
-       struct page *page;
-
-       lru_add_drain();
-       start = 0;
-       rcu_read_lock();
-
-       radix_tree_for_each_slot(slot, &mapping->i_pages, &iter, start) {
-               page = radix_tree_deref_slot(slot);
-               if (!page || radix_tree_exception(page)) {
-                       if (radix_tree_deref_retry(page)) {
-                               slot = radix_tree_iter_retry(&iter);
-                               continue;
-                       }
-               } else if (page_count(page) - page_mapcount(page) > 1) {
-                       xa_lock_irq(&mapping->i_pages);
-                       radix_tree_tag_set(&mapping->i_pages, iter.index,
-                                          MEMFD_TAG_PINNED);
-                       xa_unlock_irq(&mapping->i_pages);
-               }
-
-               if (need_resched()) {
-                       slot = radix_tree_iter_resume(slot, &iter);
-                       cond_resched_rcu();
-               }
-       }
-       rcu_read_unlock();
-}
-
-/*
- * Setting SEAL_WRITE requires us to verify there's no pending writer. However,
- * via get_user_pages(), drivers might have some pending I/O without any active
- * user-space mappings (eg., direct-IO, AIO). Therefore, we look at all pages
- * and see whether it has an elevated ref-count. If so, we tag them and wait for
- * them to be dropped.
- * The caller must guarantee that no new user will acquire writable references
- * to those pages to avoid races.
- */
-static int memfd_wait_for_pins(struct address_space *mapping)
-{
-       struct radix_tree_iter iter;
-       void __rcu **slot;
-       pgoff_t start;
-       struct page *page;
-       int error, scan;
-
-       memfd_tag_pins(mapping);
-
-       error = 0;
-       for (scan = 0; scan <= LAST_SCAN; scan++) {
-               if (!radix_tree_tagged(&mapping->i_pages, MEMFD_TAG_PINNED))
-                       break;
-
-               if (!scan)
-                       lru_add_drain_all();
-               else if (schedule_timeout_killable((HZ << scan) / 200))
-                       scan = LAST_SCAN;
-
-               start = 0;
-               rcu_read_lock();
-               radix_tree_for_each_tagged(slot, &mapping->i_pages, &iter,
-                                          start, MEMFD_TAG_PINNED) {
-
-                       page = radix_tree_deref_slot(slot);
-                       if (radix_tree_exception(page)) {
-                               if (radix_tree_deref_retry(page)) {
-                                       slot = radix_tree_iter_retry(&iter);
-                                       continue;
-                               }
-
-                               page = NULL;
-                       }
-
-                       if (page &&
-                           page_count(page) - page_mapcount(page) != 1) {
-                               if (scan < LAST_SCAN)
-                                       goto continue_resched;
-
-                               /*
-                                * On the last scan, we clean up all those tags
-                                * we inserted; but make a note that we still
-                                * found pages pinned.
-                                */
-                               error = -EBUSY;
-                       }
-
-                       xa_lock_irq(&mapping->i_pages);
-                       radix_tree_tag_clear(&mapping->i_pages,
-                                            iter.index, MEMFD_TAG_PINNED);
-                       xa_unlock_irq(&mapping->i_pages);
-continue_resched:
-                       if (need_resched()) {
-                               slot = radix_tree_iter_resume(slot, &iter);
-                               cond_resched_rcu();
-                       }
-               }
-               rcu_read_unlock();
-       }
-
-       return error;
-}
-
-static unsigned int *memfd_file_seals_ptr(struct file *file)
-{
-       if (shmem_file(file))
-               return &SHMEM_I(file_inode(file))->seals;
-
-#ifdef CONFIG_HUGETLBFS
-       if (is_file_hugepages(file))
-               return &HUGETLBFS_I(file_inode(file))->seals;
-#endif
-
-       return NULL;
-}
-
-#define F_ALL_SEALS (F_SEAL_SEAL | \
-                    F_SEAL_SHRINK | \
-                    F_SEAL_GROW | \
-                    F_SEAL_WRITE)
-
-static int memfd_add_seals(struct file *file, unsigned int seals)
-{
-       struct inode *inode = file_inode(file);
-       unsigned int *file_seals;
-       int error;
-
-       /*
-        * SEALING
-        * Sealing allows multiple parties to share a tmpfs or hugetlbfs file
-        * but restrict access to a specific subset of file operations. Seals
-        * can only be added, but never removed. This way, mutually untrusted
-        * parties can share common memory regions with a well-defined policy.
-        * A malicious peer can thus never perform unwanted operations on a
-        * shared object.
-        *
-        * Seals are only supported on special tmpfs or hugetlbfs files and
-        * always affect the whole underlying inode. Once a seal is set, it
-        * may prevent some kinds of access to the file. Currently, the
-        * following seals are defined:
-        *   SEAL_SEAL: Prevent further seals from being set on this file
-        *   SEAL_SHRINK: Prevent the file from shrinking
-        *   SEAL_GROW: Prevent the file from growing
-        *   SEAL_WRITE: Prevent write access to the file
-        *
-        * As we don't require any trust relationship between two parties, we
-        * must prevent seals from being removed. Therefore, sealing a file
-        * only adds a given set of seals to the file, it never touches
-        * existing seals. Furthermore, the "setting seals"-operation can be
-        * sealed itself, which basically prevents any further seal from being
-        * added.
-        *
-        * Semantics of sealing are only defined on volatile files. Only
-        * anonymous tmpfs and hugetlbfs files support sealing. More
-        * importantly, seals are never written to disk. Therefore, there's
-        * no plan to support it on other file types.
-        */
-
-       if (!(file->f_mode & FMODE_WRITE))
-               return -EPERM;
-       if (seals & ~(unsigned int)F_ALL_SEALS)
-               return -EINVAL;
-
-       inode_lock(inode);
-
-       file_seals = memfd_file_seals_ptr(file);
-       if (!file_seals) {
-               error = -EINVAL;
-               goto unlock;
-       }
-
-       if (*file_seals & F_SEAL_SEAL) {
-               error = -EPERM;
-               goto unlock;
-       }
-
-       if ((seals & F_SEAL_WRITE) && !(*file_seals & F_SEAL_WRITE)) {
-               error = mapping_deny_writable(file->f_mapping);
-               if (error)
-                       goto unlock;
-
-               error = memfd_wait_for_pins(file->f_mapping);
-               if (error) {
-                       mapping_allow_writable(file->f_mapping);
-                       goto unlock;
-               }
-       }
-
-       *file_seals |= seals;
-       error = 0;
-
-unlock:
-       inode_unlock(inode);
-       return error;
-}
-
-static int memfd_get_seals(struct file *file)
-{
-       unsigned int *seals = memfd_file_seals_ptr(file);
-
-       return seals ? *seals : -EINVAL;
-}
-
-long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
-{
-       long error;
-
-       switch (cmd) {
-       case F_ADD_SEALS:
-               /* disallow upper 32bit */
-               if (arg > UINT_MAX)
-                       return -EINVAL;
-
-               error = memfd_add_seals(file, arg);
-               break;
-       case F_GET_SEALS:
-               error = memfd_get_seals(file);
-               break;
-       default:
-               error = -EINVAL;
-               break;
-       }
-
-       return error;
-}
-
 static long shmem_fallocate(struct file *file, int mode, loff_t offset,
                                                         loff_t len)
 {
@@ -3677,93 +3440,6 @@ static int shmem_show_options(struct seq_file *seq, struct dentry *root)
        return 0;
 }
 
-#define MFD_NAME_PREFIX "memfd:"
-#define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
-#define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
-
-#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB)
-
-SYSCALL_DEFINE2(memfd_create,
-               const char __user *, uname,
-               unsigned int, flags)
-{
-       unsigned int *file_seals;
-       struct file *file;
-       int fd, error;
-       char *name;
-       long len;
-
-       if (!(flags & MFD_HUGETLB)) {
-               if (flags & ~(unsigned int)MFD_ALL_FLAGS)
-                       return -EINVAL;
-       } else {
-               /* Allow huge page size encoding in flags. */
-               if (flags & ~(unsigned int)(MFD_ALL_FLAGS |
-                               (MFD_HUGE_MASK << MFD_HUGE_SHIFT)))
-                       return -EINVAL;
-       }
-
-       /* length includes terminating zero */
-       len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
-       if (len <= 0)
-               return -EFAULT;
-       if (len > MFD_NAME_MAX_LEN + 1)
-               return -EINVAL;
-
-       name = kmalloc(len + MFD_NAME_PREFIX_LEN, GFP_KERNEL);
-       if (!name)
-               return -ENOMEM;
-
-       strcpy(name, MFD_NAME_PREFIX);
-       if (copy_from_user(&name[MFD_NAME_PREFIX_LEN], uname, len)) {
-               error = -EFAULT;
-               goto err_name;
-       }
-
-       /* terminating-zero may have changed after strnlen_user() returned */
-       if (name[len + MFD_NAME_PREFIX_LEN - 1]) {
-               error = -EFAULT;
-               goto err_name;
-       }
-
-       fd = get_unused_fd_flags((flags & MFD_CLOEXEC) ? O_CLOEXEC : 0);
-       if (fd < 0) {
-               error = fd;
-               goto err_name;
-       }
-
-       if (flags & MFD_HUGETLB) {
-               struct user_struct *user = NULL;
-
-               file = hugetlb_file_setup(name, 0, VM_NORESERVE, &user,
-                                       HUGETLB_ANONHUGE_INODE,
-                                       (flags >> MFD_HUGE_SHIFT) &
-                                       MFD_HUGE_MASK);
-       } else
-               file = shmem_file_setup(name, 0, VM_NORESERVE);
-       if (IS_ERR(file)) {
-               error = PTR_ERR(file);
-               goto err_fd;
-       }
-       file->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
-       file->f_flags |= O_RDWR | O_LARGEFILE;
-
-       if (flags & MFD_ALLOW_SEALING) {
-               file_seals = memfd_file_seals_ptr(file);
-               *file_seals &= ~F_SEAL_SEAL;
-       }
-
-       fd_install(fd, file);
-       kfree(name);
-       return fd;
-
-err_fd:
-       put_unused_fd(fd);
-err_name:
-       kfree(name);
-       return error;
-}
-
 #endif /* CONFIG_TMPFS */
 
 static void shmem_put_super(struct super_block *sb)