tmpfs.txt 5.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124
  1. Tmpfs is a file system which keeps all files in virtual memory.
  2. Everything in tmpfs is temporary in the sense that no files will be
  3. created on your hard drive. If you unmount a tmpfs instance,
  4. everything stored therein is lost.
  5. tmpfs puts everything into the kernel internal caches and grows and
  6. shrinks to accommodate the files it contains and is able to swap
  7. unneeded pages out to swap space. It has maximum size limits which can
  8. be adjusted on the fly via 'mount -o remount ...'
  9. If you compare it to ramfs (which was the template to create tmpfs)
  10. you gain swapping and limit checking. Another similar thing is the RAM
  11. disk (/dev/ram*), which simulates a fixed size hard disk in physical
  12. RAM, where you have to create an ordinary filesystem on top. Ramdisks
  13. cannot swap and you do not have the possibility to resize them.
  14. Since tmpfs lives completely in the page cache and on swap, all tmpfs
  15. pages currently in memory will show up as cached. It will not show up
  16. as shared or something like that. Further on you can check the actual
  17. RAM+swap use of a tmpfs instance with df(1) and du(1).
  18. tmpfs has the following uses:
  19. 1) There is always a kernel internal mount which you will not see at
  20. all. This is used for shared anonymous mappings and SYSV shared
  21. memory.
  22. This mount does not depend on CONFIG_TMPFS. If CONFIG_TMPFS is not
  23. set, the user visible part of tmpfs is not build. But the internal
  24. mechanisms are always present.
  25. 2) glibc 2.2 and above expects tmpfs to be mounted at /dev/shm for
  26. POSIX shared memory (shm_open, shm_unlink). Adding the following
  27. line to /etc/fstab should take care of this:
  28. tmpfs /dev/shm tmpfs defaults 0 0
  29. Remember to create the directory that you intend to mount tmpfs on
  30. if necessary.
  31. This mount is _not_ needed for SYSV shared memory. The internal
  32. mount is used for that. (In the 2.3 kernel versions it was
  33. necessary to mount the predecessor of tmpfs (shm fs) to use SYSV
  34. shared memory)
  35. 3) Some people (including me) find it very convenient to mount it
  36. e.g. on /tmp and /var/tmp and have a big swap partition. And now
  37. loop mounts of tmpfs files do work, so mkinitrd shipped by most
  38. distributions should succeed with a tmpfs /tmp.
  39. 4) And probably a lot more I do not know about :-)
  40. tmpfs has three mount options for sizing:
  41. size: The limit of allocated bytes for this tmpfs instance. The
  42. default is half of your physical RAM without swap. If you
  43. oversize your tmpfs instances the machine will deadlock
  44. since the OOM handler will not be able to free that memory.
  45. nr_blocks: The same as size, but in blocks of PAGE_CACHE_SIZE.
  46. nr_inodes: The maximum number of inodes for this instance. The default
  47. is half of the number of your physical RAM pages, or (on a
  48. machine with highmem) the number of lowmem RAM pages,
  49. whichever is the lower.
  50. These parameters accept a suffix k, m or g for kilo, mega and giga and
  51. can be changed on remount. The size parameter also accepts a suffix %
  52. to limit this tmpfs instance to that percentage of your physical RAM:
  53. the default, when neither size nor nr_blocks is specified, is size=50%
  54. If nr_blocks=0 (or size=0), blocks will not be limited in that instance;
  55. if nr_inodes=0, inodes will not be limited. It is generally unwise to
  56. mount with such options, since it allows any user with write access to
  57. use up all the memory on the machine; but enhances the scalability of
  58. that instance in a system with many cpus making intensive use of it.
  59. tmpfs has a mount option to set the NUMA memory allocation policy for
  60. all files in that instance (if CONFIG_NUMA is enabled) - which can be
  61. adjusted on the fly via 'mount -o remount ...'
  62. mpol=default prefers to allocate memory from the local node
  63. mpol=prefer:Node prefers to allocate memory from the given Node
  64. mpol=bind:NodeList allocates memory only from nodes in NodeList
  65. mpol=interleave prefers to allocate from each node in turn
  66. mpol=interleave:NodeList allocates from each node of NodeList in turn
  67. NodeList format is a comma-separated list of decimal numbers and ranges,
  68. a range being two hyphen-separated decimal numbers, the smallest and
  69. largest node numbers in the range. For example, mpol=bind:0-3,5,7,9-15
  70. Note that trying to mount a tmpfs with an mpol option will fail if the
  71. running kernel does not support NUMA; and will fail if its nodelist
  72. specifies a node >= MAX_NUMNODES. If your system relies on that tmpfs
  73. being mounted, but from time to time runs a kernel built without NUMA
  74. capability (perhaps a safe recovery kernel), or configured to support
  75. fewer nodes, then it is advisable to omit the mpol option from automatic
  76. mount options. It can be added later, when the tmpfs is already mounted
  77. on MountPoint, by 'mount -o remount,mpol=Policy:NodeList MountPoint'.
  78. To specify the initial root directory you can use the following mount
  79. options:
  80. mode: The permissions as an octal number
  81. uid: The user id
  82. gid: The group id
  83. These options do not have any effect on remount. You can change these
  84. parameters with chmod(1), chown(1) and chgrp(1) on a mounted filesystem.
  85. So 'mount -t tmpfs -o size=10G,nr_inodes=10k,mode=700 tmpfs /mytmpfs'
  86. will give you tmpfs instance on /mytmpfs which can allocate 10GB
  87. RAM/SWAP in 10240 inodes and it is only accessible by root.
  88. Author:
  89. Christoph Rohland <cr@sap.com>, 1.12.01
  90. Updated:
  91. Hugh Dickins <hugh@veritas.com>, 19 February 2006