xref: /freebsd/share/man/man4/md.4 (revision 3cbb4cc200f8a0ad7ed08233425ea54524a21f1c)
1.\" ----------------------------------------------------------------------------
2.\" "THE BEER-WARE LICENSE" (Revision 42):
3.\" <phk@FreeBSD.org> wrote this file.  As long as you retain this notice you
4.\" can do whatever you want with this stuff. If we meet some day, and you think
5.\" this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
6.\" ----------------------------------------------------------------------------
7.\"
8.\" $FreeBSD$
9.\"
10.Dd January 8, 2020
11.Dt MD 4
12.Os
13.Sh NAME
14.Nm md
15.Nd memory disk
16.Sh SYNOPSIS
17To compile this driver into the kernel,
18place the following lines in your
19kernel configuration file:
20.Bd -ragged -offset indent
21.Cd "device md"
22.Ed
23.Pp
24Alternatively, to load the driver as a
25module at boot time, place the following line in
26.Xr loader.conf 5 :
27.Bd -literal -offset indent
28geom_md_load="YES"
29.Ed
30.Sh DESCRIPTION
31The
32.Nm
33driver provides support for four kinds of memory backed virtual disks:
34.Bl -tag -width preload
35.It Cm malloc
36Backing store is allocated using
37.Xr malloc 9 .
38Only one malloc-bucket is used, which means that all
39.Nm
40devices with
41.Cm malloc
42backing must share the malloc-per-bucket-quota.
43The exact size of this quota varies, in particular with the amount
44of RAM in the
45system.
46The exact value can be determined with
47.Xr vmstat 8 .
48.It Cm preload
49A module loaded by
50.Xr loader 8
51with type
52.Sq md_image
53is used for backing store.
54For backwards compatibility the type
55.Sq mfs_root
56is also recognized.
57See the description of module loading directives in
58.Xr loader.conf 5
59and note that the module name will either be an absolute path to the image file
60or the name of a file in the
61.Va module_path .
62.Pp
63If the kernel is created with option
64.Dv MD_ROOT
65the first preloaded image found will become the root file system.
66.It Cm vnode
67A regular file is used as backing store.
68This allows for mounting ISO images without the tedious
69detour over actual physical media.
70.It Cm swap
71Backing store is allocated from buffer memory.
72Pages get pushed out to the swap when the system is under memory
73pressure, otherwise they stay in the operating memory.
74Using
75.Cm swap
76backing is generally preferable over
77.Cm malloc
78backing.
79.El
80.Pp
81For more information, please see
82.Xr mdconfig 8 .
83.Sh EXAMPLES
84To create a kernel with a ramdisk or MD file system, your kernel config
85needs the following options:
86.Bd -literal -offset indent
87options 	MD_ROOT			# MD is a potential root device
88options 	MD_ROOT_READONLY	# disallow mounting root writeable
89options 	MD_ROOT_SIZE=8192	# 8MB ram disk
90makeoptions	MFS_IMAGE=/h/foo/ARM-MD
91options 	ROOTDEVNAME=\\"ufs:md0\\"
92.Ed
93.Pp
94The image in
95.Pa /h/foo/ARM-MD
96will be loaded as the initial image each boot.
97To create the image to use, please follow the steps to create a file-backed
98disk found in the
99.Xr mdconfig 8
100man page.
101Other tools will also create these images, such as NanoBSD.
102.Sh ARM KERNEL OPTIONS
103On armv6 and armv7 architectures, an MD_ROOT image larger than
104approximately 55 MiB may require building a custom kernel using
105several tuning options related to kernel memory usage.
106.Bl -tag -width indent
107.It Cd options LOCORE_MAP_MB=<num>
108This configures how much memory is mapped for the kernel during
109the early initialization stages.
110The value must be at least as large as the kernel plus all preloaded
111modules, including the root image.
112There is no downside to setting this value too large, as long
113as it does not exceed the amount of physical memory.
114The default is 64 MiB.
115.It Cd options NKPT2PG=<num>
116This configures the number of kernel L2 page table pages to
117preallocate during kernel initialization.
118Each L2 page can map 4 MiB of kernel space.
119The value must be large enough to map the kernel plus all preloaded
120modules, including the root image.
121The default value is 32, which is sufficient to map 128 MiB.
122.It Cd options VM_KMEM_SIZE_SCALE=<num>
123This configures the amount of kernel virtual address (KVA) space to
124dedicate to the kmem_arena map.
125The scale value is the ratio of physical to virtual pages.
126The default value of 3 allocates a page of KVA for each 3 pages
127of physical ram in the system.
128
129The kernel and modules, including the root image, also consume KVA.
130The combination of a large root image and the default scaling
131may preallocate so much KVA that there is not enough
132remaining address space to allocate kernel stacks, IO buffers,
133and other resources that are not part of kmem_arena.
134Overallocating kmem_arena space is likely to manifest as failure to
135launch userland processes with "cannot allocate kernel stack" messages.
136
137Setting the scale value too high may result in kernel failure to allocate
138memory because kmem_arena is too small, and the failure may require
139significant runtime to manifest.
140Empirically, a value of 5 works well for a 200 MiB root image on
141a system with 2 GiB of physical ram.
142.El
143.Sh SEE ALSO
144.Xr gpart 8 ,
145.Xr loader 8 ,
146.Xr mdconfig 8 ,
147.Xr mdmfs 8 ,
148.Xr newfs 8 ,
149.Xr vmstat 8
150.Sh HISTORY
151The
152.Nm
153driver first appeared in
154.Fx 4.0
155as a cleaner replacement
156for the MFS functionality previously used in
157.Tn PicoBSD
158and in the
159.Fx
160installation process.
161.Pp
162The
163.Nm
164driver did a hostile takeover of the
165.Xr vn 4
166driver in
167.Fx 5.0 .
168.Sh AUTHORS
169The
170.Nm
171driver was written by
172.An Poul-Henning Kamp Aq Mt phk@FreeBSD.org .
173