1 =======================
2 Direct Access for files
3 =======================
8 The page cache is usually used to buffer reads and writes to files.
9 It is also used to provide the pages which are mapped into userspace
12 For block devices that are memory-like, the page cache pages would be
13 unnecessary copies of the original storage. The `DAX` code removes the
14 extra copy by performing reads and writes directly to the storage device.
15 For file mappings, the storage device is mapped directly into userspace.
21 If you have a block device which supports `DAX`, you can make a filesystem
22 on it as usual. The `DAX` code currently only supports files with a block
23 size equal to your kernel's `PAGE_SIZE`, so you may need to specify a block
24 size when creating the filesystem.
26 Currently 3 filesystems support `DAX`: ext2, ext4 and xfs. Enabling `DAX` on them
32 When mounting the filesystem, use the ``-o dax`` option on the command line or
33 add 'dax' to the options in ``/etc/fstab``. This works to enable `DAX` on all files
34 within the filesystem. It is equivalent to the ``-o dax=always`` behavior below.
37 Enabling DAX on xfs and ext4
38 ----------------------------
43 1. There exists an in-kernel file access mode flag `S_DAX` that corresponds to
44 the statx flag `STATX_ATTR_DAX`. See the manpage for statx(2) for details
45 about this access mode.
47 2. There exists a persistent flag `FS_XFLAG_DAX` that can be applied to regular
48 files and directories. This advisory flag can be set or cleared at any
49 time, but doing so does not immediately affect the `S_DAX` state.
51 3. If the persistent `FS_XFLAG_DAX` flag is set on a directory, this flag will
52 be inherited by all regular files and subdirectories that are subsequently
53 created in this directory. Files and subdirectories that exist at the time
54 this flag is set or cleared on the parent directory are not modified by
55 this modification of the parent directory.
57 4. There exist dax mount options which can override `FS_XFLAG_DAX` in the
58 setting of the `S_DAX` flag. Given underlying storage which supports `DAX` the
61 ``-o dax=inode`` means "follow `FS_XFLAG_DAX`" and is the default.
63 ``-o dax=never`` means "never set `S_DAX`, ignore `FS_XFLAG_DAX`."
65 ``-o dax=always`` means "always set `S_DAX` ignore `FS_XFLAG_DAX`."
67 ``-o dax`` is a legacy option which is an alias for ``dax=always``.
71 The option ``-o dax`` may be removed in the future so ``-o dax=always`` is
72 the preferred method for specifying this behavior.
76 Modifications to and the inheritance behavior of `FS_XFLAG_DAX` remain
77 the same even when the filesystem is mounted with a dax option. However,
78 in-core inode state (`S_DAX`) will be overridden until the filesystem is
79 remounted with dax=inode and the inode is evicted from kernel memory.
81 5. The `S_DAX` policy can be changed via:
83 a) Setting the parent directory `FS_XFLAG_DAX` as needed before files are
86 b) Setting the appropriate dax="foo" mount option
88 c) Changing the `FS_XFLAG_DAX` flag on existing regular files and
89 directories. This has runtime constraints and limitations that are
90 described in 6) below.
92 6. When changing the `S_DAX` policy via toggling the persistent `FS_XFLAG_DAX`
93 flag, the change to existing regular files won't take effect until the
94 files are closed by all processes.
100 There are 2 per-file dax flags. One is a persistent inode setting (`FS_XFLAG_DAX`)
101 and the other is a volatile flag indicating the active state of the feature
104 `FS_XFLAG_DAX` is preserved within the filesystem. This persistent config
105 setting can be set, cleared and/or queried using the `FS_IOC_FS`[`GS`]`ETXATTR` ioctl
106 (see ioctl_xfs_fsgetxattr(2)) or an utility such as 'xfs_io'.
108 New files and directories automatically inherit `FS_XFLAG_DAX` from
109 their parent directory **when created**. Therefore, setting `FS_XFLAG_DAX` at
110 directory creation time can be used to set a default behavior for an entire
113 To clarify inheritance, here are 3 examples:
117 .. code-block:: shell
120 xfs_io -c 'chattr +x' a
124 ------[outcome]------
131 .. code-block:: shell
134 xfs_io -c 'chattr +x' a
137 ------[outcome]------
144 .. code-block:: shell
147 xfs_io -c 'chattr +x' c
150 ------[outcome]------
155 The current enabled state (`S_DAX`) is set when a file inode is instantiated in
156 memory by the kernel. It is set based on the underlying media support, the
157 value of `FS_XFLAG_DAX` and the filesystem's dax mount option.
159 statx can be used to query `S_DAX`.
163 That only regular files will ever have `S_DAX` set and therefore statx
164 will never indicate that `S_DAX` is set on directories.
166 Setting the `FS_XFLAG_DAX` flag (specifically or through inheritance) occurs even
167 if the underlying media does not support dax and/or the filesystem is
168 overridden with a mount option.
171 Implementation Tips for Block Driver Writers
172 --------------------------------------------
174 To support `DAX` in your block driver, implement the 'direct_access'
175 block device operation. It is used to translate the sector number
176 (expressed in units of 512-byte sectors) to a page frame number (pfn)
177 that identifies the physical page for the memory. It also returns a
178 kernel virtual address that can be used to access the memory.
180 The direct_access method takes a 'size' parameter that indicates the
181 number of bytes being requested. The function should return the number
182 of bytes that can be contiguously accessed at that offset. It may also
183 return a negative errno if an error occurs.
185 In order to support this method, the storage must be byte-accessible by
186 the CPU at all times. If your device uses paging techniques to expose
187 a large amount of memory through a smaller window, then you cannot
188 implement direct_access. Equally, if your device can occasionally
189 stall the CPU for an extended period, you should also not attempt to
190 implement direct_access.
192 These block devices may be used for inspiration:
193 - brd: RAM backed block device driver
194 - dcssblk: s390 dcss block device driver
195 - pmem: NVDIMM persistent memory driver
198 Implementation Tips for Filesystem Writers
199 ------------------------------------------
201 Filesystem support consists of:
203 * Adding support to mark inodes as being `DAX` by setting the `S_DAX` flag in
205 * Implementing ->read_iter and ->write_iter operations which use
206 :c:func:`dax_iomap_rw()` when inode has `S_DAX` flag set
207 * Implementing an mmap file operation for `DAX` files which sets the
208 `VM_MIXEDMAP` and `VM_HUGEPAGE` flags on the `VMA`, and setting the vm_ops to
209 include handlers for fault, pmd_fault, page_mkwrite, pfn_mkwrite. These
210 handlers should probably call :c:func:`dax_iomap_fault()` passing the
211 appropriate fault size and iomap operations.
212 * Calling :c:func:`iomap_zero_range()` passing appropriate iomap operations
213 instead of :c:func:`block_truncate_page()` for `DAX` files
214 * Ensuring that there is sufficient locking between reads, writes,
215 truncates and page faults
217 The iomap handlers for allocating blocks must make sure that allocated blocks
218 are zeroed out and converted to written extents before being returned to avoid
219 exposure of uninitialized data through mmap.
221 These filesystems may be used for inspiration:
225 ext2: see Documentation/filesystems/ext2.rst
229 xfs: see Documentation/admin-guide/xfs.rst
233 ext4: see Documentation/filesystems/ext4/
236 Handling Media Errors
237 ---------------------
239 The libnvdimm subsystem stores a record of known media error locations for
240 each pmem block device (in gendisk->badblocks). If we fault at such location,
241 or one with a latent error not yet discovered, the application can expect
242 to receive a `SIGBUS`. Libnvdimm also allows clearing of these errors by simply
243 writing the affected sectors (through the pmem driver, and if the underlying
244 NVDIMM supports the clear_poison DSM defined by ACPI).
246 Since `DAX` IO normally doesn't go through the ``driver/bio`` path, applications or
247 sysadmins have an option to restore the lost data from a prior ``backup/inbuilt``
248 redundancy in the following ways:
250 1. Delete the affected file, and restore from a backup (sysadmin route):
251 This will free the filesystem blocks that were being used by the file,
252 and the next time they're allocated, they will be zeroed first, which
253 happens through the driver, and will clear bad sectors.
255 2. Truncate or hole-punch the part of the file that has a bad-block (at least
256 an entire aligned sector has to be hole-punched, but not necessarily an
257 entire filesystem block).
259 These are the two basic paths that allow `DAX` filesystems to continue operating
260 in the presence of media errors. More robust error recovery mechanisms can be
261 built on top of this in the future, for example, involving redundancy/mirroring
262 provided at the block layer through DM, or additionally, at the filesystem
263 level. These would have to rely on the above two tenets, that error clearing
264 can happen either by sending an IO through the driver, or zeroing (also through
271 Even if the kernel or its modules are stored on a filesystem that supports
272 `DAX` on a block device that supports `DAX`, they will still be copied into RAM.
274 The DAX code does not work correctly on architectures which have virtually
275 mapped caches such as ARM, MIPS and SPARC.
277 Calling :c:func:`get_user_pages()` on a range of user memory that has been
278 mmaped from a `DAX` file will fail when there are no 'struct page' to describe
279 those pages. This problem has been addressed in some device drivers
280 by adding optional struct page support for pages under the control of
281 the driver (see `CONFIG_NVDIMM_PFN` in ``drivers/nvdimm`` for an example of
282 how to do this). In the non struct page cases `O_DIRECT` reads/writes to
283 those memory ranges from a non-`DAX` file will fail
288 `O_DIRECT` reads/writes _of a `DAX` file do work, it is the memory that
289 is being accessed that is key here). Other things that will not work in
290 the non struct page case include RDMA, :c:func:`sendfile()` and