1 .. SPDX-License-Identifier: GPL-2.0
7 Tmpfs is a file system which keeps all of its files in virtual memory.
10 Everything in tmpfs is temporary in the sense that no files will be
11 created on your hard drive. If you unmount a tmpfs instance,
12 everything stored therein is lost.
14 tmpfs puts everything into the kernel internal caches and grows and
15 shrinks to accommodate the files it contains and is able to swap
16 unneeded pages out to swap space, if swap was enabled for the tmpfs
17 mount. tmpfs also supports THP.
19 tmpfs extends ramfs with a few userspace configurable options listed and
20 explained further below, some of which can be reconfigured dynamically on the
21 fly using a remount ('mount -o remount ...') of the filesystem. A tmpfs
22 filesystem can be resized but it cannot be resized to a size below its current
23 usage. tmpfs also supports POSIX ACLs, and extended attributes for the
24 trusted.* and security.* namespaces. ramfs does not use swap and you cannot
25 modify any parameter for a ramfs filesystem. The size limit of a ramfs
26 filesystem is how much memory you have available, and so care must be taken if
27 used so to not run out of memory.
29 An alternative to tmpfs and ramfs is to use brd to create RAM disks
30 (/dev/ram*), which allows you to simulate a block device disk in physical RAM.
31 To write data you would just then need to create an regular filesystem on top
32 this ramdisk. As with ramfs, brd ramdisks cannot swap. brd ramdisks are also
33 configured in size at initialization and you cannot dynamically resize them.
34 Contrary to brd ramdisks, tmpfs has its own filesystem, it does not rely on the
37 Since tmpfs lives completely in the page cache and optionally on swap,
38 all tmpfs pages will be shown as "Shmem" in /proc/meminfo and "Shared" in
39 free(1). Notice that these counters also include shared memory
40 (shmem, see ipcs(1)). The most reliable way to get the count is
41 using df(1) and du(1).
43 tmpfs has the following uses:
45 1) There is always a kernel internal mount which you will not see at
46 all. This is used for shared anonymous mappings and SYSV shared
49 This mount does not depend on CONFIG_TMPFS. If CONFIG_TMPFS is not
50 set, the user visible part of tmpfs is not built. But the internal
51 mechanisms are always present.
53 2) glibc 2.2 and above expects tmpfs to be mounted at /dev/shm for
54 POSIX shared memory (shm_open, shm_unlink). Adding the following
55 line to /etc/fstab should take care of this::
57 tmpfs /dev/shm tmpfs defaults 0 0
59 Remember to create the directory that you intend to mount tmpfs on
62 This mount is _not_ needed for SYSV shared memory. The internal
63 mount is used for that. (In the 2.3 kernel versions it was
64 necessary to mount the predecessor of tmpfs (shm fs) to use SYSV
67 3) Some people (including me) find it very convenient to mount it
68 e.g. on /tmp and /var/tmp and have a big swap partition. And now
69 loop mounts of tmpfs files do work, so mkinitrd shipped by most
70 distributions should succeed with a tmpfs /tmp.
72 4) And probably a lot more I do not know about :-)
75 tmpfs has three mount options for sizing:
77 ========= ============================================================
78 size The limit of allocated bytes for this tmpfs instance. The
79 default is half of your physical RAM without swap. If you
80 oversize your tmpfs instances the machine will deadlock
81 since the OOM handler will not be able to free that memory.
82 nr_blocks The same as size, but in blocks of PAGE_SIZE.
83 nr_inodes The maximum number of inodes for this instance. The default
84 is half of the number of your physical RAM pages, or (on a
85 machine with highmem) the number of lowmem RAM pages,
86 whichever is the lower.
87 noswap Disables swap. Remounts must respect the original settings.
88 By default swap is enabled.
89 ========= ============================================================
91 These parameters accept a suffix k, m or g for kilo, mega and giga and
92 can be changed on remount. The size parameter also accepts a suffix %
93 to limit this tmpfs instance to that percentage of your physical RAM:
94 the default, when neither size nor nr_blocks is specified, is size=50%
96 If nr_blocks=0 (or size=0), blocks will not be limited in that instance;
97 if nr_inodes=0, inodes will not be limited. It is generally unwise to
98 mount with such options, since it allows any user with write access to
99 use up all the memory on the machine; but enhances the scalability of
100 that instance in a system with many CPUs making intensive use of it.
102 tmpfs also supports Transparent Huge Pages which requires a kernel
103 configured with CONFIG_TRANSPARENT_HUGEPAGE and with huge supported for
104 your system (has_transparent_hugepage(), which is architecture specific).
105 The mount options for this are:
107 ====== ============================================================
108 huge=0 never: disables huge pages for the mount
109 huge=1 always: enables huge pages for the mount
110 huge=2 within_size: only allocate huge pages if the page will be
111 fully within i_size, also respect fadvise()/madvise() hints.
112 huge=3 advise: only allocate huge pages if requested with
114 ====== ============================================================
116 There is a sysfs file which you can also use to control system wide THP
117 configuration for all tmpfs mounts, the file is:
119 /sys/kernel/mm/transparent_hugepage/shmem_enabled
121 This sysfs file is placed on top of THP sysfs directory and so is registered
122 by THP code. It is however only used to control all tmpfs mounts with one
123 single knob. Since it controls all tmpfs mounts it should only be used either
124 for emergency or testing purposes. The values you can set for shmem_enabled are:
126 == ============================================================
127 -1 deny: disables huge on shm_mnt and all mounts, for
129 -2 force: enables huge on shm_mnt and all mounts, w/o needing
131 == ============================================================
133 tmpfs has a mount option to set the NUMA memory allocation policy for
134 all files in that instance (if CONFIG_NUMA is enabled) - which can be
135 adjusted on the fly via 'mount -o remount ...'
137 ======================== ==============================================
138 mpol=default use the process allocation policy
139 (see set_mempolicy(2))
140 mpol=prefer:Node prefers to allocate memory from the given Node
141 mpol=bind:NodeList allocates memory only from nodes in NodeList
142 mpol=interleave prefers to allocate from each node in turn
143 mpol=interleave:NodeList allocates from each node of NodeList in turn
144 mpol=local prefers to allocate memory from the local node
145 ======================== ==============================================
147 NodeList format is a comma-separated list of decimal numbers and ranges,
148 a range being two hyphen-separated decimal numbers, the smallest and
149 largest node numbers in the range. For example, mpol=bind:0-3,5,7,9-15
151 A memory policy with a valid NodeList will be saved, as specified, for
152 use at file creation time. When a task allocates a file in the file
153 system, the mount option memory policy will be applied with a NodeList,
154 if any, modified by the calling task's cpuset constraints
155 [See Documentation/admin-guide/cgroup-v1/cpusets.rst] and any optional flags,
156 listed below. If the resulting NodeLists is the empty set, the effective
157 memory policy for the file will revert to "default" policy.
159 NUMA memory allocation policies have optional flags that can be used in
160 conjunction with their modes. These optional flags can be specified
161 when tmpfs is mounted by appending them to the mode before the NodeList.
162 See Documentation/admin-guide/mm/numa_memory_policy.rst for a list of
163 all available memory allocation policy mode flags and their effect on
168 =static is equivalent to MPOL_F_STATIC_NODES
169 =relative is equivalent to MPOL_F_RELATIVE_NODES
171 For example, mpol=bind=static:NodeList, is the equivalent of an
172 allocation policy of MPOL_BIND | MPOL_F_STATIC_NODES.
174 Note that trying to mount a tmpfs with an mpol option will fail if the
175 running kernel does not support NUMA; and will fail if its nodelist
176 specifies a node which is not online. If your system relies on that
177 tmpfs being mounted, but from time to time runs a kernel built without
178 NUMA capability (perhaps a safe recovery kernel), or with fewer nodes
179 online, then it is advisable to omit the mpol option from automatic
180 mount options. It can be added later, when the tmpfs is already mounted
181 on MountPoint, by 'mount -o remount,mpol=Policy:NodeList MountPoint'.
184 To specify the initial root directory you can use the following mount
187 ==== ==================================
188 mode The permissions as an octal number
191 ==== ==================================
193 These options do not have any effect on remount. You can change these
194 parameters with chmod(1), chown(1) and chgrp(1) on a mounted filesystem.
197 tmpfs has a mount option to select whether it will wrap at 32- or 64-bit inode
200 ======= ========================
201 inode64 Use 64-bit inode numbers
202 inode32 Use 32-bit inode numbers
203 ======= ========================
205 On a 32-bit kernel, inode32 is implicit, and inode64 is refused at mount time.
206 On a 64-bit kernel, CONFIG_TMPFS_INODE64 sets the default. inode64 avoids the
207 possibility of multiple files with the same inode number on a single device;
208 but risks glibc failing with EOVERFLOW once 33-bit inode numbers are reached -
209 if a long-lived tmpfs is accessed by 32-bit applications so ancient that
210 opening a file larger than 2GiB fails with EINVAL.
213 So 'mount -t tmpfs -o size=10G,nr_inodes=10k,mode=700 tmpfs /mytmpfs'
214 will give you tmpfs instance on /mytmpfs which can allocate 10GB
215 RAM/SWAP in 10240 inodes and it is only accessible by root.
219 Christoph Rohland <cr@sap.com>, 1.12.01
221 Hugh Dickins, 4 June 2007
223 KOSAKI Motohiro, 16 Mar 2010
225 Chris Down, 13 July 2020