Previous Chapter - Table of Contents - Next Chapter
- Explain concepts such as inodes, directory files, and extended attributes.
- Create and format filesystems.
- Check and fix errors on filesystems.
- Mount and unmount filesystems.
- Use network file systems.
- Understand how to automount filesystems and make sure they mount at boot.
Extended Attributes associate metadata not interpreted directly by filesystem with files. Four namespaces exist:
- user
- trusted
- security
- system
System namespace used for Access Control Lists (ACLs) and security namespace used by SELinux.
Flag values stored in file inode, maybe modified and set only by root user. Viewed with lsattr and set with chattr. Flags may be set for files:
i: Immutable. File with immutable attribute cannot be modified (not even my root). Cannot be deleted or renamed. No hard link can be created to it, and no data can be written to file. Only superuser can set or clear this attribute.a: Append-only. FIle with append-only attribute set can only be opened in append mode for writing. Only superuser can set ot clear this attribute.d: No-dump. File with no-dump attribute set is ignored when the dump program is run. Useful for swap and cache files that you don't want to waste time backing up.A: No atime update. File with no-atime-update attribute set will not modify its atime (access time) record when file accessed but not otherwise modified. Can increase performance on some systems because it reduces amount of disk I/O on system.
Note: there are other flags that can be set; typing man chattr will show whole list. Format for chattr:
$ chattr [+|-|=mode] filenamelsattr used to display attributes for file:
$ lsattr filenameEvery filesystem type has utility for formatting (making) filesystem on partition. Generic name for these utilities: mkfs. However, this is just frontend for filesystem-specific programs, each of which may have particular options.
Thus, following two commands entirely equivalent:
$ sudo mkfs -t ext4 /dev/sda10
$ sudo mkfs.ext4 /dev/sda10General format for mkfs:
mkfs [-t fstype] [options] [device-file]where [device-file] is usually device name like /dev/sda3 of /dev/vg/lvm1.
Each filesystem has own particular options that can be set when formatting. Eg. when creating ext4 filesystem, journalling settings = one thing to keep in mind. Include defining journal file size, whether or not to use external journal file.
Should look at man page for each of mkfs.* programs to see details.
Every filesystem type has utility designed to check for errors (and hopefully fix any that are found). Generic name for these utilities: fsck. However, just frontend for filesystem-specific programs.
Thus, following two commands entirely equivalent:
$ sudo fsck -t ext4 /dev/sda10
$ sudo fsck.ext4 /dev/sda10If filesystem is of type understood by operating system, can almost always just do:
$ sudo fsck /dev/sda10and system will figure out type by examining first few bytes on partition.
fsck run automatically after set number of mounts, or set interval since last time it was run, or after abnormal shutdown. Should only be run on unmounted filesystems. Can force a check of all ounted filesystems at boot by doing:
$ sudo touch /forcefsck
$ sudo rebootThe file /forcefsck will disappear after successful check. One reason this is valuable trick: can do fsck on root filesystem, which is hard to do on running system.
General format for fsck:
fsck [-t fstype] [options] [device-file]where [device-file] is usually device name like /dev/sda3 or /dev/vg/lvm1. Usually, do not need to specify filesystem type, as fsck can figure it out by examining superblocks at start of partition.
Can control whether any errors found should be fixed one by one manually with -r option, or automatically, as best possible, by using -a option, etc. In addition, each filesystem type may have own particular options that can be set when checking.
Note: journalling filesystems much faster to check than older generation filesystems for two reasons:
- Rarely need to scan entire partition for errors, as everything but very last transaction logged and confirmed, so takes almost no time to check
- Even if whole filesystem checked, newer filesystems designed with fast fsck in mind. Older filesystems did not think much about this when designed as sizes were much smaller
Should look at man page for each of fsck.* programs to see details.
All accessible files in Linux organized into one large hierarchical tree structure with head of tree being root directory (/). However, common to have more than one partition (each of which can have own filesystem type) joined together in same filesystem tree. Partitions can also be on different physical devices, even on network.
mount program allows attaching at any point in tree structure. umount allows detaching them.
Mount point is directory where filesystem attached. Must exist before mount can use it. mkdir can be used to create expty directory. If pre-existing directory used + contains files prior to being used as mount point, files will be hidden after mounting. These files are not deleted and will again be visible when filesystem unmounted.
By default, only superuser can mount/unmount filesystems.
Each filesystem mounted under specific directory:
$ sudo mount -t ext4 /dev/sdb4 home- Mounts ext4 filesystem
- Usually not necessary to specify type with
-toption - Filesystem is located on specific partition of hard drive (
/dev/sdb4) - Filesystem mounted at position
/homein current directory tree - Any files residing in original
/homedirectory hidden until partition unmounted
General form for mount:
mount [options] <source> <directory>Note: in this example, filesystem mounted by using device node it resides on. However, also possible to mount using label or UUID. Thus, following all equivalent:
$ sudo mount /dev/sda2 /home
$ sudo mount LABEL=home /home
$ sudo mount -L home /home
$ sudo mount UUID=26d58ee2-9d20-4dc7-b6ab-aa87c3cfb69a /home
$ sudo mount -U 26d58ee2-9d20-4dc7-b6ab-aa87c3cfb69a /homeLabels assigned by filesystem type specific utilities, such as e2label, and UUIDs assigned when partitions created as containers for filesystem, formatted with mkfs.
While any of these three methods for specifying device can be used, modern systems deprecate using device node form because names can change according to how the system is booted, which hard drives found first, etc. Labels = improvement, but on rare occasions, could have two partitions that wind up with same label. UUIDs, however, should always be unique + created when partitions created.
mount takes many options, some generic like -a (mount all filesystems mentioned in /etc/fstab) and many filesystem specific. Has very long man page. Common example:
$ sudo mount -o remount,ro /myfswhich remounts filesystem with read-only attribute.
An exampe of hot to get quick summary of mount options:
$ mount --help
Filesystems can be unmounted:
$ umount [device-file | mount-point]Below, some examples of how to unmount filesystem:
- Unmount
/homefilesystem:$ sudo umount /home
- Unmount the
/dev/sda3device:$ sudo umount /dev/sda3
Note: command to unmount filesystem is umount (not unmount!).
Like mount, umount has many options, many of which specific to filesystem type. man pages = best source for specific option information.
Most common error encountered when unmounting filesystem: trying to do this on filesystem currently in use, ie. current applications using files or other entries in filesystem.
Can be as simple as having terminal window open in directory on mounted filesystem. Just using cd in that window, or killing it, will get rid of device is busy error and allow unmounting.
However, if other processes inducing this error, must kill those processes before unmounting filesystem. Can use fuser to find out which users using filesystem and kill them (be careful with this, may also want to warn users first). Can also use lsof ("list open files") to try and see which files being used and blocking unmounting.
Common to mount remote filesystems through network shares so they appear as if they were on local machine. Probably most common method used historically: NFS (Network File System).
NFS originally developed by Sun Microsystems in 1989, has been continuously updated. Modern systems use NFSv4, which has been continuously updated since 2000.
Other network filesystems include AFS (Andrew File System), and SMB (Server Message Book), also termed CIFS (Common Internet File System).
Because network filesystem may be unavailable at any time, either because it is not present on network share, or the network is unavailable, systems have to be prepared for this possibility.
Thus, in such circumstances, system should be instructed not to get hung, or blocked, while waiting longer than specified period. Can be specified in mount command:
$ sudo mount -t nfs myserver.com:/shdir /mnt/shdiror in /etc/fstab. Put following line in /etc/fstab to mount on boot or with mount -a:
myserver.com:/shdir /mnt/shdir nfs rsize=8192,wsize=8192,timeo=14,intr 0 0System may try to mount NFS filesystem before network is up. netdev and noauto options can be used. For more information, check man nfs, examine mount options.
Can also be solved using autofs or automount.
Mount has large amount of options, some of which specific to nfs. See man pages for both nfs and mount for details.
During system initialization, the command mount -a executed. Mounts all filesystems listed in /etc/fstab configuration file. Entries can refer to both local + remote network-mounted filesystems. Below shows example of how to mount all filesystems listed in /etc/fstab configuration file during system boot.
File shows what filesystems may be automatically mounted at boot, and where they may be found on the local machine or network. Can specify who may mount them and with what permissions, and other relevant options. Some lines refer to special pseudo-filesystems such as proc, sys, devpts.
Each record in /etc/fstab file contains white space separated fields of information about filesystem to be mounted:
-
Device node, label, or UUID
For filesystems which do not have device node, such as tmpfs, proc, and sysfs, this field is just placeholder. Sometimes will see the word none in that column, or used on command line.
-
Mount point
Can also be placeholder, like for swap, which is not mounted anywhere.
-
Filesystem type (ie. ext4, xfs, btrfs, vfat)
-
A comma-separated list of options
-
dump frequency (or a 0)
Used by rarely used
dump -wcommand. -
fsck pass number (or 0, meaning do not check state at boot)
mount and umount utilities can use information in /etc/fstab. In such case, could type:
$ sudo mount /usr/srcinstead of
$ sudo mount LABEL=src /usr/srcin above example.
Linux systems have long had ability to mount filesystem only when filesystem needed. Historically, done using autofs. This utility requires installation of autofs package using appropriate package manager + configuration of files in /etc.
While autofs very flaxible + well understood, systemd-based systems (including all recent enterprise Linux distributions) come with automount facilities built into systemd framework. Configuring this = simple as adding a line in /etc/fstab specifying proper device, mount point, mount options:
LABEL=Sam128 /SAM ext4 noauto,x-systemd.automount,x-systemd.device-timeout=10,x-systemd.idle-timeout=30 0 0and then, either rebooting or issuing command:
$ sudo systemctl daemon-reload
$ sudo systemctl restart local-fs.targetNext, will give example and explain options.
Example provided next mounts USB pen drive that is always plugged into system, only when it is used. Options in /etc/fstab:
noauto: do not mount at boot. Here,autodoes not refer to automountx-systemd.automount: use the systemd automount facilityx-systemd.automount.device-timeout=10: if this device is not available, say it is a network device accessible through NFS, timeout after 10 seconds instead of getting hungx-systemd.automount.idle-timeout=30: if device is not used for 30 seconds, unmount it
Note: device may be mounted during boot, but then should be unmounted after timeout specified. Below shows how device is only available one it is used.
List of currently mounted filesystems can be seen by typing:
