Chinaunix首页 | 论坛 | 博客
  • 博客访问: 2190329
  • 博文数量: 609
  • 博客积分: 10061
  • 博客等级: 上将
  • 技术积分: 5920
  • 用 户 组: 普通用户
  • 注册时间: 2008-06-25 08:30
文章分类

全部博文(609)

文章存档

2010年(13)

2009年(39)

2008年(558)

我的朋友

分类: LINUX

2008-07-04 22:46:54

 Tmpfs简介

 Tmpfs is a file system which keeps all files in virtual memory.
   
 Everything in tmpfs is temporary in the sense that no files will be
 created on your hard drive. If you unmount a tmpfs instance,
 everything stored therein is lost.
  
 tmpfs puts everything into the kernel internal caches and grows and
 shrinks to accommodate the files it contains and is able to swap
 unneeded pages out to swap space. It has maximum size limits which can
 be adjusted on the fly via 'mount -o remount ...'
 
 If you compare it to ramfs (which was the template to create tmpfs)
 you gain swapping and limit checking. Another similar thing is the RAM
 disk (/dev/ram*), which simulates a fixed size hard disk in physical
 RAM, where you have to create an ordinary filesystem on top. Ramdisks
 cannot swap and you do not have the possibility to resize them.
 
 Since tmpfs lives completely in the page cache and on swap, all tmpfs
 pages currently in memory will show up as cached. It will not show up
 as shared or something like that. Further on you can check the actual
 RAM+swap use of a tmpfs instance with df(1) and du(1).
 
 
 tmpfs has the following uses:
 
 1) There is always a kernel internal mount which you will not see at
    all. This is used for shared anonymous mappings and SYSV shared
    memory.
 
    This mount does not depend on CONFIG_TMPFS. If CONFIG_TMPFS is not
    set, the user visible part of tmpfs is not build. But the internal
    mechanisms are always present.
 
 2) glibc 2.2 and above expects tmpfs to be mounted at /dev/shm for
    POSIX shared memory (shm_open, shm_unlink). Adding the following
    line to /etc/fstab should take care of this:
 
         tmpfs   /dev/shm        tmpfs   defaults        0 0
 
    Remember to create the directory that you intend to mount tmpfs on
    if necessary.
 
    This mount is _not_ needed for SYSV shared memory. The internal
    mount is used for that. (In the 2.3 kernel versions it was
    necessary to mount the predecessor of tmpfs (shm fs) to use SYSV
    shared memory)
 
 3) Some people (including me) find it very convenient to mount it
    e.g. on /tmp and /var/tmp and have a big swap partition. And now
    loop mounts of tmpfs files do work, so mkinitrd shipped by most
    distributions should succeed with a tmpfs /tmp.
 
 4) And probably a lot more I do not know about :-)
 
 
 tmpfs has three mount options for sizing:
 
 size:      The limit of allocated bytes for this tmpfs instance. The
            default is half of your physical RAM without swap. If you
            oversize your tmpfs instances the machine will deadlock
            since the OOM handler will not be able to free that memory.
 nr_blocks: The same as size, but in blocks of PAGE_CACHE_SIZE.
 nr_inodes: The maximum number of inodes for this instance. The default
            is half of the number of your physical RAM pages, or (on a
            machine with highmem) the number of lowmem RAM pages,
            whichever is the lower.
 
 These parameters accept a suffix k, m or g for kilo, mega and giga and
 can be changed on remount.  The size parameter also accepts a suffix %
 to limit this tmpfs instance to that percentage of your physical RAM:
 the default, when neither size nor nr_blocks is specified, is size=50%
 
 If nr_blocks=0 (or size=0), blocks will not be limited in that instance;
 if nr_inodes=0, inodes will not be limited.  It is generally unwise to
 mount with such options, since it allows any user with write access to
 use up all the memory on the machine; but enhances the scalability of
 that instance in a system with many cpus making intensive use of it.
 
 
 tmpfs has a mount option to set the NUMA memory allocation policy for
 all files in that instance (if CONFIG_NUMA is enabled) - which can be
 adjusted on the fly via 'mount -o remount ...'
 
 mpol=default             prefers to allocate memory from the local node
 mpol=prefer:Node         prefers to allocate memory from the given Node
 mpol=bind:NodeList       allocates memory only from nodes in NodeList
 mpol=interleave          prefers to allocate from each node in turn
 mpol=interleave:NodeList allocates from each node of NodeList in turn
 
 NodeList format is a comma-separated list of decimal numbers and ranges,
 a range being two hyphen-separated decimal numbers, the smallest and
 largest node numbers in the range.  For example, mpol=bind:0-3,5,7,9-15
 
 Note that trying to mount a tmpfs with an mpol option will fail if the
 running kernel does not support NUMA; and will fail if its nodelist
 specifies a node which is not online.  If your system relies on that
 tmpfs being mounted, but from time to time runs a kernel built without
 NUMA capability (perhaps a safe recovery kernel), or with fewer nodes
 online, then it is advisable to omit the mpol option from automatic
 mount options.  It can be added later, when the tmpfs is already mounted
 on MountPoint, by 'mount -o remount,mpol=Policy:NodeList MountPoint'.
 
 
 To specify the initial root directory you can use the following mount
 options:
 
 mode:   The permissions as an octal number
 uid:    The user id
 gid:    The group id
 
 These options do not have any effect on remount. You can change these
 parameters with chmod(1), chown(1) and chgrp(1) on a mounted filesystem.
 
 
 So 'mount -t tmpfs -o size=10G,nr_inodes=10k,mode=700 tmpfs /mytmpfs'
 will give you tmpfs instance on /mytmpfs which can allocate 10GB
 RAM/SWAP in 10240 inodes and it is only accessible by root.


 Author:
    Christoph Rohland , 1.12.01
 Updated:
    Hugh Dickins , 4 June 2007
 
 
阅读(1640) | 评论(1) | 转发(0) |
给主人留下些什么吧!~~

chinaunix网友2008-08-05 13:23:26

SAP99,支持下,也欢迎访问我的博客, SAP资料多多 http://sap99.cublog.cn PA系列教材 web_dynpro_for_java Web_Dynpro_for_abap abap资料-在SAP中查询TABL SAP实施顾问宝典 SAP配置(录像)-维护供应 SAP配置(录像)-定义税务 sap全面概述中文