kernel disk storage API
struct disk *
The disk storage API permits kernel device drivers providing access to disk-like
storage devices to advertise the device to other kernel components, including
Each disk device is described by a struct disk
structure, which contains a variety of parameters for the disk device,
function pointers for various methods that may be performed on the device, as
well as private data storage for the device driver. In addition, some fields
are reserved for use by GEOM in managing access to the device and its
GEOM has the ownership of struct disk
drivers must allocate storage for it with the
() function, fill in the fields
() when the device is
ready to service requests.
adds an alias for the disk and must be called before
(), but may be called multiple
times. For each alias added, a device node will be created with
in the same way primary device nodes are created with
. Care should be taken to ensure that
only one driver creates aliases for any given name.
() can be called by the driver
after modifying d_mediasize
to notify GEOM
about the disk capacity change. The flags
field should be set to either M_WAITOK, or M_NOWAIT.
() orphans all of the providers
associated with the drive, setting an error condition of ENXIO in each one. In
addition, it prevents a re-taste on last close for writing if an error
condition has been set in the provider. After calling
(), the device driver is not
allowed to access the contents of struct disk
() function takes a second
, which must always be
. If GEOM detects that
the driver is compiled against an unsupported version, it will ignore the
device and print a warning on the console.
The following fields identify the disk device described by the structure
instance, and must be filled in prior to submitting the structure to
() and may not be subsequently
- Optional flags indicating to the storage framework what optional features
or descriptions the storage device driver supports. Currently supported
DISKFLAG_OPEN (maintained by
(maintained by device driver), and
DISKFLAG_CANFLUSHCACHE (maintained by
- const char *
- Holds the name of the storage device class, e.g.,
ahd”. This value typically uniquely
identifies a particular driver device, and must not conflict with devices
serviced by other device drivers.
- Holds the instance of the storage device class, e.g.,
4”. This namespace is managed by
the device driver, and assignment of unit numbers might be a property of
probe order, or in some cases topology. Together, the
d_unit values will uniquely identify a
disk storage device.
The following fields identify various disk device methods, if implemented:
- disk_open_t *
- Optional: invoked when the disk device is opened. If no method is
provided, open will always succeed.
- disk_close_t *
- Optional: invoked when the disk device is closed. Although an error code
may be returned, the call should always terminate any state setup by the
corresponding open method call.
- disk_strategy_t *
- Mandatory: invoked when a new struct bio
is to be initiated on the disk device.
- disk_ioctl_t *
- Optional: invoked when an I/O control operation is initiated on the disk
device. Please note that for security reasons these operations should not
be able to affect other devices than the one on which they are
- dumper_t *
- Optional: if configured with
this function is invoked from a very restricted system state after a
kernel panic to record a copy of the system RAM to the disk.
- disk_getattr_t *
- Optional: if this method is provided, it gives the disk driver the
opportunity to override the default GEOM response to BIO_GETATTR requests.
This function should return -1 if the attribute is not handled, 0 if the
attribute is handled, or an errno to be passed to g_io_deliver().
- disk_gone_t *
- Optional: if this method is provided, it will be called after disk_gone()
is called, once GEOM has finished its cleanup process. Once this callback
is called, it is safe for the disk driver to free all of its resources, as
it will not be receiving further calls from GEOM.
The following fields identify the size and granularity of the disk device. These
fields must stay stable from return of the drivers open method until the close
method is called, but it is perfectly legal to modify them in the open method
- The sector size of the disk device in bytes.
- The size of the disk device in bytes.
- The maximum supported size in bytes of an I/O request. Requests larger
than this size will be chopped up by GEOM.
These optional fields can provide extra information about the disk device. Do
not initialize these fields if the field/concept does not apply. These fields
must stay stable from return of the drivers open method until the close method
is called, but it is perfectly legal to modify them in the open method before
- The number of sectors and heads advertised on the disk device by the
firmware or BIOS. These values are almost universally bogus, but on some
architectures necessary for the correct calculation of disk
- These two fields can be used to describe the width and location of natural
performance boundaries for most disk technologies. Please see
src/sys/geom/notes for details.
- This field can and should be used to store disk's serial number if the
d_getattr method described above isn't implemented, or if it does not
support the GEOM::ident attribute.
- This field can be used to store the disk vendor and product
- This field can be used to store the PCI vendor ID for the HBA connected to
- This field can be used to store the PCI device ID for the HBA connected to
- This field can be used to store the PCI subvendor ID for the HBA connected
to the disk.
- This field can be used to store the PCI subdevice ID for the HBA connected
to the disk.
This field may be used by the device driver to store a pointer to private data
to implement the disk service.
- void *
- Private data pointer. Typically used to store a pointer to the drivers
softc structure for this disk
This manual page was written by Robert
Disk aliases are not a general purpose aliasing mechanism, but are intended only
to ease the transition from one name to another. They can be used to ensure
that nvd0 and nda0 are the same thing. They cannot be used to implement the
diskX concept from macOS.