xref: /freebsd/share/man/man9/disk.9 (revision 4b9d6057)
1.\"
2.\" Copyright (c) 2003 Robert N. M. Watson
3.\" All rights reserved.
4.\"
5.\" Redistribution and use in source and binary forms, with or without
6.\" modification, are permitted provided that the following conditions
7.\" are met:
8.\" 1. Redistributions of source code must retain the above copyright
9.\"    notice(s), this list of conditions and the following disclaimer as
10.\"    the first lines of this file unmodified other than the possible
11.\"    addition of one or more copyright notices.
12.\" 2. Redistributions in binary form must reproduce the above copyright
13.\"    notice(s), this list of conditions and the following disclaimer in the
14.\"    documentation and/or other materials provided with the distribution.
15.\"
16.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY
17.\" EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
18.\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
19.\" DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY
20.\" DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21.\" (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
22.\" SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
23.\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
26.\" DAMAGE.
27.\"
28.Dd April 30, 2020
29.Dt DISK 9
30.Os
31.Sh NAME
32.Nm disk
33.Nd kernel disk storage API
34.Sh SYNOPSIS
35.In geom/geom_disk.h
36.Ft struct disk *
37.Fn disk_alloc void
38.Ft void
39.Fn disk_create "struct disk *disk" "int version"
40.Ft void
41.Fn disk_gone "struct disk *disk"
42.Ft void
43.Fn disk_destroy "struct disk *disk"
44.Ft int
45.Fn disk_resize "struct disk *disk" "int flags"
46.Ft void
47.Fn disk_add_alias "struct disk *disk" "const char *alias"
48.Sh DESCRIPTION
49The disk storage API permits kernel device drivers providing access to
50disk-like storage devices to advertise the device to other kernel
51components, including
52.Xr GEOM 4
53and
54.Xr devfs 5 .
55.Pp
56Each disk device is described by a
57.Vt "struct disk"
58structure, which contains a variety of parameters for the disk device,
59function pointers for various methods that may be performed on the device,
60as well as private data storage for the device driver.
61In addition, some fields are reserved for use by GEOM in managing access
62to the device and its statistics.
63.Pp
64GEOM has the ownership of
65.Vt "struct disk" ,
66and drivers must allocate storage for it with the
67.Fn disk_alloc
68function,
69fill in the fields and call
70.Fn disk_create
71when the device is ready to service requests.
72.Fn disk_add_alias
73adds an alias for the disk and must be called before
74.Fn disk_create ,
75but may be called multiple times.
76For each alias added, a device node will be created with
77.Xr make_dev_alias 9
78in the same way primary device nodes are created with
79.Xr make_dev 9
80for
81.Va d_name
82and
83.Va d_unit .
84Care should be taken to ensure that only one driver creates aliases
85for any given name.
86.Fn disk_resize
87can be called by the driver after modifying
88.Va d_mediasize
89to notify GEOM about the disk capacity change.
90The
91.Fa flags
92field should be set to either M_WAITOK, or M_NOWAIT.
93.Fn disk_gone
94orphans all of the providers associated with the drive, setting an error
95condition of ENXIO in each one.
96In addition, it prevents a re-taste on last close for writing if an error
97condition has been set in the provider.
98After calling
99.Fn disk_destroy ,
100the device driver is not allowed to access the contents of
101.Vt "struct disk"
102anymore.
103.Pp
104The
105.Fn disk_create
106function
107takes a second parameter,
108.Fa version ,
109which must always be passed
110.Dv DISK_VERSION .
111If GEOM detects that the driver is compiled against an unsupported version,
112it will ignore the device and print a warning on the console.
113.Ss Descriptive Fields
114The following fields identify the disk device described by the structure
115instance, and must be filled in prior to submitting the structure to
116.Fn disk_create
117and may not be subsequently changed:
118.Bl -tag -width indent
119.It Vt u_int Va d_flags
120Optional flags indicating to the storage framework what optional features
121or descriptions the storage device driver supports.
122Currently supported flags are
123.Dv DISKFLAG_OPEN
124(maintained by storage framework),
125.Dv DISKFLAG_CANDELETE
126(maintained by device driver),
127and
128.Dv DISKFLAG_CANFLUSHCACHE
129(maintained by device driver).
130.It Vt "const char *" Va d_name
131Holds the name of the storage device class, e.g.,
132.Dq Li ahd .
133This value typically uniquely identifies a particular driver device,
134and must not conflict with devices serviced by other device drivers.
135.It Vt u_int Va d_unit
136Holds the instance of the storage device class, e.g.,
137.Dq Li 4 .
138This namespace is managed by the device driver, and assignment of unit
139numbers might be a property of probe order, or in some cases topology.
140Together, the
141.Va d_name
142and
143.Va d_unit
144values will uniquely identify a disk storage device.
145.El
146.Ss Disk Device Methods
147The following fields identify various disk device methods, if implemented:
148.Bl -tag -width indent
149.It Vt "disk_open_t *" Va d_open
150Optional: invoked when the disk device is opened.
151If no method is provided, open will always succeed.
152.It Vt "disk_close_t *" Va d_close
153Optional: invoked when the disk device is closed.
154Although an error code may be returned, the call should always terminate
155any state setup by the corresponding open method call.
156.It Vt "disk_strategy_t *" Va d_strategy
157Mandatory: invoked when a new
158.Vt "struct bio"
159is to be initiated on the disk device.
160.It Vt "disk_ioctl_t *" Va d_ioctl
161Optional: invoked when an I/O control operation is initiated on the disk device.
162Please note that for security reasons these operations should not
163be able to affect other devices than the one on which they are performed.
164.It Vt "dumper_t *" Va d_dump
165Optional: if configured with
166.Xr dumpon 8 ,
167this function is invoked from a very restricted system state after a
168kernel panic to record a copy of the system RAM to the disk.
169.It Vt "disk_getattr_t *" Va d_getattr
170Optional: if this method is provided, it gives the disk driver the
171opportunity to override the default GEOM response to BIO_GETATTR requests.
172This function should return -1 if the attribute is not handled, 0 if the
173attribute is handled, or an errno to be passed to
174.Fn g_io_deliver .
175.It Vt "disk_gone_t *" Va d_gone
176Optional: if this method is provided, it will be called after
177.Fn disk_gone
178is called, once GEOM has finished its cleanup process.
179Once this callback is called, it is safe for the disk driver to free all of
180its resources, as it will not be receiving further calls from GEOM.
181.El
182.Ss Mandatory Media Properties
183The following fields identify the size and granularity of the disk device.
184These fields must stay stable from return of the drivers open method until
185the close method is called, but it is perfectly legal to modify them in
186the open method before returning.
187.Bl -tag -width indent
188.It Vt u_int Va d_sectorsize
189The sector size of the disk device in bytes.
190.It Vt off_t Va d_mediasize
191The size of the disk device in bytes.
192.It Vt u_int Va d_maxsize
193The maximum supported size in bytes of an I/O request.
194Requests larger than this size will be chopped up by GEOM.
195.El
196.Ss Optional Media Properties
197These optional fields can provide extra information about the disk
198device.
199Do not initialize these fields if the field/concept does not apply.
200These fields must stay stable from return of the drivers open method until
201the close method is called, but it is perfectly legal to modify them in
202the open method before returning.
203.Bl -tag -width indent
204.It Vt u_int Va d_fwsectors , Vt u_int Va d_fwheads
205The number of sectors and heads advertised on the disk device by the
206firmware or BIOS.
207These values are almost universally bogus, but on some architectures
208necessary for the correct calculation of disk partitioning.
209.It Vt u_int Va d_stripeoffset , Vt u_int Va d_stripesize
210These two fields can be used to describe the width and location of
211natural performance boundaries for most disk technologies.
212Please see
213.Pa src/sys/geom/notes
214for details.
215.It Vt char Va d_ident[DISK_IDENT_SIZE]
216This field can and should be used to store disk's serial number if the
217d_getattr method described above isn't implemented, or if it does not
218support the GEOM::ident attribute.
219.It Vt char Va d_descr[DISK_IDENT_SIZE]
220This field can be used to store the disk vendor and product description.
221.It Vt uint16_t Va d_hba_vendor
222This field can be used to store the PCI vendor ID for the HBA connected to
223the disk.
224.It Vt uint16_t Va d_hba_device
225This field can be used to store the PCI device ID for the HBA connected to
226the disk.
227.It Vt uint16_t Va d_hba_subvendor
228This field can be used to store the PCI subvendor ID for the HBA connected to
229the disk.
230.It Vt uint16_t Va d_hba_subdevice
231This field can be used to store the PCI subdevice ID for the HBA connected to
232the disk.
233.El
234.Ss Driver Private Data
235This field may be used by the device driver to store a pointer to
236private data to implement the disk service.
237.Bl -tag -width indent
238.It Vt "void *" Va d_drv1
239Private data pointer.
240Typically used to store a pointer to the drivers
241.Vt softc
242structure for this disk device.
243.El
244.Sh SEE ALSO
245.Xr GEOM 4 ,
246.Xr devfs 5
247.Sh HISTORY
248The
249.Nm kernel disk storage API
250first appeared in
251.Fx 4.9 .
252.Sh AUTHORS
253This manual page was written by
254.An Robert Watson .
255.Sh BUGS
256Disk aliases are not a general purpose aliasing mechanism, but are
257intended only to ease the transition from one name to another.
258They can be used to ensure that nvd0 and nda0 are the same thing.
259They cannot be used to implement the diskX concept from macOS.
260