diff options
Diffstat (limited to 'docs/library/vfs.rst')
| -rw-r--r-- | docs/library/vfs.rst | 208 |
1 files changed, 208 insertions, 0 deletions
diff --git a/docs/library/vfs.rst b/docs/library/vfs.rst new file mode 100644 index 000000000..fcd06eb43 --- /dev/null +++ b/docs/library/vfs.rst @@ -0,0 +1,208 @@ +:mod:`vfs` -- virtual filesystem control +======================================== + +.. module:: vfs + :synopsis: virtual filesystem control + +The ``vfs`` module contains functions for creating filesystem objects and +mounting/unmounting them in the Virtual Filesystem. + +Filesystem mounting +------------------- + +Some ports provide a Virtual Filesystem (VFS) and the ability to mount multiple +"real" filesystems within this VFS. Filesystem objects can be mounted at either +the root of the VFS, or at a subdirectory that lives in the root. This allows +dynamic and flexible configuration of the filesystem that is seen by Python +programs. Ports that have this functionality provide the :func:`mount` and +:func:`umount` functions, and possibly various filesystem implementations +represented by VFS classes. + +.. function:: mount(fsobj, mount_point, *, readonly) + + Mount the filesystem object *fsobj* at the location in the VFS given by the + *mount_point* string. *fsobj* can be a a VFS object that has a ``mount()`` + method, or a block device. If it's a block device then the filesystem type + is automatically detected (an exception is raised if no filesystem was + recognised). *mount_point* may be ``'/'`` to mount *fsobj* at the root, + or ``'/<name>'`` to mount it at a subdirectory under the root. + + If *readonly* is ``True`` then the filesystem is mounted read-only. + + During the mount process the method ``mount()`` is called on the filesystem + object. + + Will raise ``OSError(EPERM)`` if *mount_point* is already mounted. + +.. function:: umount(mount_point) + + Unmount a filesystem. *mount_point* can be a string naming the mount location, + or a previously-mounted filesystem object. During the unmount process the + method ``umount()`` is called on the filesystem object. + + Will raise ``OSError(EINVAL)`` if *mount_point* is not found. + +.. class:: VfsFat(block_dev) + + Create a filesystem object that uses the FAT filesystem format. Storage of + the FAT filesystem is provided by *block_dev*. + Objects created by this constructor can be mounted using :func:`mount`. + + .. staticmethod:: mkfs(block_dev) + + Build a FAT filesystem on *block_dev*. + +.. class:: VfsLfs1(block_dev, readsize=32, progsize=32, lookahead=32) + + Create a filesystem object that uses the `littlefs v1 filesystem format`_. + Storage of the littlefs filesystem is provided by *block_dev*, which must + support the :ref:`extended interface <block-device-interface>`. + Objects created by this constructor can be mounted using :func:`mount`. + + See :ref:`filesystem` for more information. + + .. staticmethod:: mkfs(block_dev, readsize=32, progsize=32, lookahead=32) + + Build a Lfs1 filesystem on *block_dev*. + + .. note:: There are reports of littlefs v1 failing in certain situations, + for details see `littlefs issue 347`_. + +.. class:: VfsLfs2(block_dev, readsize=32, progsize=32, lookahead=32, mtime=True) + + Create a filesystem object that uses the `littlefs v2 filesystem format`_. + Storage of the littlefs filesystem is provided by *block_dev*, which must + support the :ref:`extended interface <block-device-interface>`. + Objects created by this constructor can be mounted using :func:`mount`. + + The *mtime* argument enables modification timestamps for files, stored using + littlefs attributes. This option can be disabled or enabled differently each + mount time and timestamps will only be added or updated if *mtime* is enabled, + otherwise the timestamps will remain untouched. Littlefs v2 filesystems without + timestamps will work without reformatting and timestamps will be added + transparently to existing files once they are opened for writing. When *mtime* + is enabled `os.stat` on files without timestamps will return 0 for the timestamp. + + See :ref:`filesystem` for more information. + + .. staticmethod:: mkfs(block_dev, readsize=32, progsize=32, lookahead=32) + + Build a Lfs2 filesystem on *block_dev*. + + .. note:: There are reports of littlefs v2 failing in certain situations, + for details see `littlefs issue 295`_. + +.. class:: VfsPosix(root=None) + + Create a filesystem object that accesses the host POSIX filesystem. + If *root* is specified then it should be a path in the host filesystem to use + as the root of the ``VfsPosix`` object. Otherwise the current directory of + the host filesystem is used. + +.. _littlefs v1 filesystem format: https://github.com/ARMmbed/littlefs/tree/v1 +.. _littlefs v2 filesystem format: https://github.com/ARMmbed/littlefs +.. _littlefs issue 295: https://github.com/ARMmbed/littlefs/issues/295 +.. _littlefs issue 347: https://github.com/ARMmbed/littlefs/issues/347 + +Block devices +------------- + +A block device is an object which implements the block protocol. This enables a +device to support MicroPython filesystems. The physical hardware is represented +by a user defined class. The :class:`AbstractBlockDev` class is a template for +the design of such a class: MicroPython does not actually provide that class, +but an actual block device class must implement the methods described below. + +A concrete implementation of this class will usually allow access to the +memory-like functionality of a piece of hardware (like flash memory). A block +device can be formatted to any supported filesystem and mounted using ``os`` +methods. + +See :ref:`filesystem` for example implementations of block devices using the +two variants of the block protocol described below. + +.. _block-device-interface: + +Simple and extended interface +............................. + +There are two compatible signatures for the ``readblocks`` and ``writeblocks`` +methods (see below), in order to support a variety of use cases. A given block +device may implement one form or the other, or both at the same time. The second +form (with the offset parameter) is referred to as the "extended interface". + +Some filesystems (such as littlefs) that require more control over write +operations, for example writing to sub-block regions without erasing, may require +that the block device supports the extended interface. + +.. class:: AbstractBlockDev(...) + + Construct a block device object. The parameters to the constructor are + dependent on the specific block device. + + .. method:: readblocks(block_num, buf) + readblocks(block_num, buf, offset) + + The first form reads aligned, multiples of blocks. + Starting at the block given by the index *block_num*, read blocks from + the device into *buf* (an array of bytes). + The number of blocks to read is given by the length of *buf*, + which will be a multiple of the block size. + + The second form allows reading at arbitrary locations within a block, + and arbitrary lengths. + Starting at block index *block_num*, and byte offset within that block + of *offset*, read bytes from the device into *buf* (an array of bytes). + The number of bytes to read is given by the length of *buf*. + + .. method:: writeblocks(block_num, buf) + writeblocks(block_num, buf, offset) + + The first form writes aligned, multiples of blocks, and requires that the + blocks that are written to be first erased (if necessary) by this method. + Starting at the block given by the index *block_num*, write blocks from + *buf* (an array of bytes) to the device. + The number of blocks to write is given by the length of *buf*, + which will be a multiple of the block size. + + The second form allows writing at arbitrary locations within a block, + and arbitrary lengths. Only the bytes being written should be changed, + and the caller of this method must ensure that the relevant blocks are + erased via a prior ``ioctl`` call. + Starting at block index *block_num*, and byte offset within that block + of *offset*, write bytes from *buf* (an array of bytes) to the device. + The number of bytes to write is given by the length of *buf*. + + Note that implementations must never implicitly erase blocks if the offset + argument is specified, even if it is zero. + + .. method:: ioctl(op, arg) + + Control the block device and query its parameters. The operation to + perform is given by *op* which is one of the following integers: + + - 1 -- initialise the device (*arg* is unused) + - 2 -- shutdown the device (*arg* is unused) + - 3 -- sync the device (*arg* is unused) + - 4 -- get a count of the number of blocks, should return an integer + (*arg* is unused) + - 5 -- get the number of bytes in a block, should return an integer, + or ``None`` in which case the default value of 512 is used + (*arg* is unused) + - 6 -- erase a block, *arg* is the block number to erase + + As a minimum ``ioctl(4, ...)`` must be intercepted; for littlefs + ``ioctl(6, ...)`` must also be intercepted. The need for others is + hardware dependent. + + Prior to any call to ``writeblocks(block, ...)`` littlefs issues + ``ioctl(6, block)``. This enables a device driver to erase the block + prior to a write if the hardware requires it. Alternatively a driver + might intercept ``ioctl(6, block)`` and return 0 (success). In this case + the driver assumes responsibility for detecting the need for erasure. + + Unless otherwise stated ``ioctl(op, arg)`` can return ``None``. + Consequently an implementation can ignore unused values of ``op``. Where + ``op`` is intercepted, the return value for operations 4 and 5 are as + detailed above. Other operations should return 0 on success and non-zero + for failure, with the value returned being an ``OSError`` errno code. |