Got root?

lecture in color

Got root?

Everything we have done so far required no privileges.
Could do it all as a normal user.

So far, you're still normal...

Illusion: becoming a superuser gives one `super powers'.
Reality: you can do most anything as a normal user.

Being root

can delete any file.
can work around any file protection.
has to be more careful than a normal user.
can easily kill a system with a word.

Becoming root

su: `switch user'
The command
```
 
% su
Password: 
#
```
makes you root!
- new prompt! (#)
- new shell! (/bin/sh or /bin/bash)
- can now do anything!
- even things that are really dumb!.

Stupid root tricks

Consider the command
```
rm -rf .*
```
- * matches non-dot files (misses .cshrc)
- .* matches . and ..
- rm -rf .* not only removes your directory, but all siblings of it in the directory above it!
- Correct:
```
rm -rf .??*   # match anything starting with . with >= 3 chars.
```
Consider the command
```
cp -r /home/couch /home/grouch
```
- copies all files from /home/couch to /home/grouch.
- messes up all protections and timestamps!
- Correct:
```
 
(cd /home/couch; tar cf - .) | (cd /home/grouch; tar xfBp -)
```
- tar cf - .: put all contents of current directory on standard output.
  - c: create archive
  - f -: write to standard output.
- tar xfBp -: retrieve contents from standard input,
  - x: extract
  - f -: from standard input
  - B: which is a pipe
  - p: preserve owner and mode

Ethics

When you work as root, you're subject to a higher code of ethics than a normal user.
'#' is not a reminder of power, but of responsibility.
Working as root is a matter of building trust.
Betray that trust and you'll never work as a system admin again.
See The SAGE Code of Ethics (which I personally had some small part in reviewing, being a person on the committee that formulated it)

Privacy

Rules for privacy are different for root.
normal user: what you can see is typically ok to see.
root user: you can see anything and are required to restrain yourself from looking at files you aren't authorized to see.
- property rights: proprietary software and documents
- business ethics: you should not invade privacy.
- legality: you don't have a right to know private things filed on the system.
- And you have an obligation to report what you know about illegal activities.

Power

Abuse of power is really easy for root.
normal user: can't do anything to any other user.
root user: can log people out, corrupt or destroy files, impersonate people, snoop on network, hack into other computers.

Rules of root

(from my own training for my own sysadmins)

Never do anything you don't think you understand.
- don't type things without understanding them
- study them as necessary to understand what they do before typing them.
Never do anything you can't undo.
- save copies of files you change
```
 
cp /etc/inetd.conf /etc/inetd.conf.orig
```
- Don't do things that are irreversible without warning management.
Never do anything as root that you can do as a normal user.
- to compile software: chown it to yourself, be yourself, compile it.
- to work on a subsystem: chown files, fix, chown back.
Never run untested or untrusted 3rd party software as root!
- security breaches/trojans.
- clueless programmers
Never run emacs as root! Learn vi for root editing.
- When editing a file through a symlink, vi preserves the link, emacs breaks it.
- Example: /etc/hosts -> /etc/inet/hosts
- After emacs /etc/hosts, these are different files.
- A sun in this state (with two different ideas of hosts) is horribly broken.
- (or chown the file to you, become yourself, then run emacs)
Stamp out emacs! and those @#$%^& file~'s.
- vi is available on every UNIX system.
- emacs is not typically available on new systems.
- Better yet, learn ed, the only editor that'll work in single user mode on a damaged system.
Document everything you do.
- so others know what's been done.
- so others can restore software after a system upgrade.
Coordinate changes with other admins
- so you won't undo the changes of others.
- so you won't break something someone else installed.

Root's view of the world

normal user	root user
`login`	`su`
`/bin/csh` and relatives	`/bin/sh` and relatives
directories	filesystems
programs	daemons

One example: filesystems

root can restructure how you see directories.
device: something with which the operating system can interact.
disk device: something that acts like a disk.
filesystem: a structure on a disk device that allows one to represent files and directories.

Life of a filesystem

filesystems are raw devices, with a specific format, mounted as directories.

 raw device --------> device containing -----------> 'mapped' filesystem
          newfs, mkfs    filesystem        mount

exploring filesystems

Kinds of filesystems

Local: UFS (Solaris), ext2 (linux)
Network: NFS (Sun), AFS (CMU), Samba (Linux/NT)

Investigating filesystems

df - show `mounted filesystems'.

 [couch@lin01 ~]$ df
 Filesystem           1k-blocks      Used Available Use% Mounted on
 /dev/sda1              2522048   1388580   1005352  58% /
 /dev/sda2              5779532     23380   5462568   0% /var/local
 conmoto:/export/27/loc/ap/linux-x86-libc6
                       52127512  37298448  14307792  72% /local
 conmoto:/export/27/loc
                       52127512  37298448  14307792  72% /loc
 conmoto:/export/14/p/1/comp150net
                       35244336  21586384  13305512  62% /p/1/comp150net

short for disk free

may require -k option on some systems in order to produce reasonable output:

 blackhole{couch}87: df
 /proc              (/proc             ):       0 blocks      921 files
 /                  (/dev/dsk/c0t3d0s0 ):  111508 blocks    39265 files
 /usr               (/dev/dsk/c0t3d0s6 ):  521280 blocks   346298 files
 /dev/fd            (fd                ):       0 blocks        0 files
 /var               (/dev/dsk/c0t3d0s1 ):   52200 blocks    19890 files
 /tmp               (swap              ):  145912 blocks    10452 files
 /cf                (conmoto:/export/0/cf):33121572 blocks  3394458 files
 /loc               (conmoto:/export/27/loc):29658122 blocks  5158058 files
 /u/f               (conmoto:/export/1/u/f):37855660 blocks  3809948 files
 /var/mail          (yogi:/export/mail ):25129160 blocks  1709248 files
 /local             (conmoto:/export/27/loc/SLINK):29658122 blocks  5158058 files
 blackhole{couch}88: df -k
 Filesystem            kbytes    used   avail capacity  Mounted on
 /proc                      0       0       0     0%    /proc
 /dev/dsk/c0t3d0s0      83319   27565   47423    37%    /
 /dev/dsk/c0t3d0s6     746982  486342  200882    71%    /usr
 fd                         0       0       0     0%    /dev/fd
 /dev/dsk/c0t3d0s1      41407   15307   21960    42%    /var
 swap                   73228     284   72944     1%    /tmp
 conmoto:/export/0/cf 34751145 18190359 16213275    53%    /cf
 conmoto:/export/27/loc
                      52127508 37298447 14307786    73%    /loc
 conmoto:/export/1/u/f
                      34751145 15824616 18579018    46%    /u/f
 yogi:/export/mail    13472260  907564 11880328     8%    /var/mail
 conmoto:/export/27/loc/SLINK
                      52127508 37298447 14307786    73%    /local
 blackhole{couch}89:

Filesystem: where the filesystem resides (local device or remote disk)
- /dev/dsk/c0t3d0s6 - a disk device
  - SCSI disk (pronounced `Scuzzy')
  - c0 - controller 0
  - t6 - target 6
  - d0 - disk 0
  - s6 - subchannel 6
mount point: where the filesystem appears to be within the directory hierarchy.

Studying output of `df`

two kinds of filesystems
- local: filesystem is /dev/something
```
 /dev/sda1              2522048   1388580   1005352  58% /
```
- remote: filesystem is hostname:/directory/path
```
 conmoto:/export/27/loc
                       52127512  37298448  14307792  72% /loc
```
  says that /loc here is /export/27/loc on the server |conmoto|.

File service architecture

You can't login to conmoto.
It's only a file server.
It's typical for machines to have one function:
- fileserver
- cycle server (user login)
- web server

Problem of the day #1

If allegro:/some/thing is a mount point for blackhole:/export/1 and andante:/t/some/thing is a mount point for blackhole:/export/1/thing then which directories correspond to one another (are aliases for one another) on andante and allegro?

Mounting filesystems

mount - make a filesystem appear
```
 /sbin/mount /dev/dsk/c0t3d0s1 /mnt
```
makes the contents of /dev/dsk/c0t3d0s1 become known as /mnt
- /mnt - a standard mount point for testing filesystems.
- temporary.
- present already in most UNIX's
umount - make a filesystem disappear.
```
 /sbin/umount /mnt
```
undoes what the previous command did.

Mount rules

you can't just assume that anything you want to do will work.
device mounting rules
- only one device can be mounted on one mount point (directory).
- only one device can be specified for each mount point.
- mounting a device over a directory hides existing subdirectories (that reappear when you unmount)
network mounting rules
- cannot mount any filesystem that isn't `exported'.
- cannot export a filesystem you mounted from elsewhere

True story

once I had a full disk.
df said 100% full.
du said it was 0% full (relative to the number of blocks available).
files were hidden under a mount point!
caveat: df can see under mount points; du cannot!

Making filesystems permanent

Systems differ in how they record that filesystems are permanent.

Sun: /etc/vfstab lists mount commands to do at boot:

 #device           device             mount  FS    fsck     mount    mount
 #to mount         to fsck            point  type  pass    at boot  options
 /dev/dsk/c0t3d0s3 -                  -      swap   -        no        -
 /dev/dsk/c0t3d0s0 /dev/rdsk/c0t3d0s0 /      ufs    1        no        -
 /dev/dsk/c0t3d0s6 /dev/rdsk/c0t3d0s6 /usr   ufs    1        no        -
 /dev/dsk/c0t3d0s1 /dev/rdsk/c0t3d0s1 /var   ufs    1        no        -
 ^                 ^                  ^      ^     ^^^ when to check it. 
 ^                 ^                  ^      ^^^ type of filesystem. 
 ^                 ^                  ^^^^ where to make this appear. 
 ^                 ^^^^^^^^^^^^^^^^^^ corresponding 'raw' (character) device
 ^^^^^^^^^^^^^^^^^ block device

device to mount: the block device (/dev/dsk)
device to check: the raw (character) device (/dev/rdsk)
mount point: where to make it appear in the filesystem.

Linux: /etc/fstab

 [couch@lin01 ~]$ more /etc/fstab
 # device                mount                   type    options         dump
 /dev/sda1               /                       ext2    defaults        1 1
 /dev/cdrom              /mnt/cdrom              iso9660 noauto,owner,ro 0 0
 /dev/sda2               /var/local              ext2    defaults        1 2
 /dev/fd0                /mnt/floppy             auto    noauto,owner    0 0
 none                    /proc                   proc    defaults        0 0
 none                    /dev/pts                devpts  gid=5,mode=620  0 0
 /dev/sda3               swap                    swap    defaults        0 0
 conmoto:/export/27/loc/ap/linux-x86-libc6  /local nfs hard,rsize=8192,wsize=8192 0 0
 conmoto:/export/27/loc /loc  nfs hard,rsize=8192,wsize=8192  0 0
 conmoto:/export/14/p/1/comp150net /p/1/comp150net nfs hard,rsize=8192,wsize=8192 0 0
 [couch@lin01 ~]$

in both cases, the file represents
- mount points and targets
- types of filesystems
- mount options
  - ro - 'read-only'
  - noauto - don't mount this at boot

Automounting

To make things more confusing, some mounting occurs on demand when you cd to a particular directory.
So far, only used on Solaris in our department.
Can tell what it's doing by examining /etc/auto_* and NIS+ maps auto_*.org_dir

Allegro's automount table (assignment 3)

niscat auto_master.org_dir defines submaps.

 blackhole{couch}73: niscat auto_master.org_dir
 /nfs auto_nfs
 /u auto_u
 /- auto_direct
 /home auto_home

niscat auto_direct.org_dir defines `direct mounts'

 blackhole{couch}74: niscat auto_direct.org_dir
 /var/mail -vers=3,-actimeo=0    yogi:/export/mail
 /local conmoto:/export/27/loc/SLINK
 /r conmoto:/export/0/r
 /p conmoto:/export/14/p
 /cf conmoto:/export/0/cf
 /g conmoto:/export/0/g
 /loc conmoto:/export/27/loc
 /s conmoto:/export/28/s
 /patches conmoto:/export/0/patches
 blackhole{couch}75:

niscat auto_u.org_dir mounts things in /u:

 blackhole{couch}75: niscat auto_u.org_dir
 j conmoto:/export/1/u/j
 f conmoto:/export/1/u/f
 O conmoto:/export/28/u/O
 s conmoto:/export/28/u/s 
 m conmoto:/export/28/u/m
 t conmoto:/export/1/u/t 
 o conmoto:/export/28/u/o 
 blackhole{couch}76:

niscat auto_loc.org_dir mounts things in /loc

 blackhole{couch}76: niscat auto_loc.org_dir
 constraint conmoto:/export/0/loc/&
 SLINK conmoto:/export/0/loc/&
 TT_DB conmoto:/export/0/loc/&
 X11 conmoto:/export/0/loc/&
 adm conmoto:/export/0/loc/&
 aladdin conmoto:/export/0/loc/&
 alone conmoto:/export/0/loc/&
 case conmoto:/export/0/loc/&
 dbms conmoto:/export/0/loc/&
 emacs-18.59 conmoto:/export/0/loc/&
 games conmoto:/export/0/loc/&
 gnu conmoto:/export/0/loc/&
 graphics conmoto:/export/0/loc/&
 interleaf conmoto:/export/0/loc/&
 lang conmoto:/export/0/loc/&
 mail conmoto:/export/0/loc/&
 net conmoto:/export/0/loc/&
 libs conmoto:/export/0/loc/&
 pkg conmoto:/export/0/loc/&
 publish conmoto:/export/0/loc/&
 shells conmoto:/export/0/loc/&
 slink conmoto:/export/0/loc/&
 utils conmoto:/export/0/loc/&
 xclients conmoto:/export/0/loc/&
 ee26 conmoto:/export/0/loc/&
 encryption conmoto:/export/0/loc/&
 fax conmoto:/export/0/loc/&
 from conmoto:/export/0/loc/&
 g conmoto:/export/0/loc/&
 tree conmoto:/export/0/loc/&
 out conmoto:/export/0/loc/&
 ncd conmoto:/export/0/loc/&
 parallel conmoto:/export/0/loc/&
 tftpboot conmoto:/export/0/loc/&
 bsd-4.3 conmoto:/export/0/loc/&
 cad conmoto:/export/0/loc/&
 ce conmoto:/export/0/loc/&
 ee128 conmoto:/export/0/loc/&
 eesof-6.0 conmoto:/export/0/loc/&
 espresso conmoto:/export/0/loc/&
 idl conmoto:/export/0/loc/&
 motif-1.2 conmoto:/export/0/loc/&
 motif-examples conmoto:/export/0/loc/&
 motif-uimx conmoto:/export/0/loc/&
 sofa conmoto:/export/0/loc/&
 packages conmoto:/export/0/loc/&
 acct conmoto:/export/0/loc/&
 editors conmoto:/export/0/loc/&
 kerberos conmoto:/export/0/loc/&
 kirby conmoto:/export/0/loc/&
 lang2 conmoto:/export/0/loc/&
 sound conmoto:/export/0/loc/&
 vis conmoto:/export/0/loc/&
 www conmoto:/export/0/loc/&
 mentor conmoto:/export/0/loc/&
 adm-MOVED conmoto:/export/0/loc/&
 att conmoto:/export/0/loc/&
 avanti conmoto:/export/0/loc/&
 ansoft conmoto:/export/0/loc/&
 chinese conmoto:/export/0/loc/&
 apache conmoto:/export/0/loc/&
 ap conmoto:/export/0/loc/&
 postgres conmoto:/export/0/loc/&
 graphics2 conmoto:/export/0/loc/&
 jserv conmoto:/export/0/loc/&
 3com conmoto:/export/0/loc/&
 java conmoto:/export/0/loc/&
 geom conmoto:/export/0/loc/&
 sgi0 conmoto:/export/0/loc/&
 gut conmoto:/export/0/loc/&
 cadence conmoto:/export/0/loc/&
 gnu2 conmoto:/export/0/loc/&
 fs conmoto:/export/0/loc/&
 rep conmoto:/export/0/loc/&
 blackhole{couch}77:

Special syntax:
```
 geom conmoto:/export/0/loc/&
```
means that conmoto:/export/0/loc/geom maps to /loc/geom

Digging deeper: making a filesystem

Partition the disk appropriately.
Make a filesystem on the disk.
Mount the filesystem within your directory structure.
Copy files into it if necessary.

Raw disk geometry

Cylinder: one stripe of data circumnavigating the whole physical disk (utilizing multiple Tracks/Heads)
Track/Head: one media reading implement on a disk.
Sector: a part of a track, usually 512 bytes of data.
Cylinders * Tracks/Cylinder * Sectors/Track * Bytes/Sector = Raw Bytes.
The triple (Cylinders, Heads(or Tracks), Sectors) is called the geometry of a disk.
For modern disks, this is fiction. Geometry is variable from outside to inside cylinders (more path length means higher density).

Disk concepts:

disk: a physical piece of hardware.
partition: a subdivision of the disk (a contiguous region of blocks).
filesystem: a logical structure for a partition containing inodes and blocks.
mounting: the process of mapping a filesystem into a directory hierarchy.

Partitioning under solaris

Under Sun Solaris, use the format command:

 # format
 Searching for disks...done
 
 
 AVAILABLE DISK SELECTIONS:
        0. c0t2d0 <MICROP-1598-15MD1052405-AS24 cyl 1911 alt 2 hd 15 sec 71>
           /iommu@0,10000000/sbus@0,10001000/espdma@5,8400000/esp@5,8800000/sd@2,0
        1. c0t3d0 <SUN1.05 cyl 2036 alt 2 hd 14 sec 72>
           /iommu@0,10000000/sbus@0,10001000/espdma@5,8400000/esp@5,8800000/sd@3,0
 Specify disk (enter its number): 0
 selecting c0t2d0
 [disk formatted]
 
 
 FORMAT MENU:
         disk       - select a disk
         type       - select (define) a disk type
         partition  - select (define) a partition table
         current    - describe the current disk
         format     - format and analyze the disk
         repair     - repair a defective sector
         label      - write label to the disk
         analyze    - surface analysis
         defect     - defect list management
         backup     - search for backup labels
         verify     - read and display labels
         save       - save new disk/partition definitions
         inquiry    - show vendor, product and revision
         volname    - set 8-character volume name
         !<cmd>     - execute <cmd>, then return
         quit
 format> part
 
 
 PARTITION MENU:
         0      - change `0' partition
         1      - change `1' partition
         2      - change `2' partition
         3      - change `3' partition
         4      - change `4' partition
         5      - change `5' partition
         6      - change `6' partition
         7      - change `7' partition
         select - select a predefined table
         modify - modify a predefined partition table
         name   - name the current table
         print  - display the current table
         label  - write partition map and label to the disk
         !<cmd> - execute <cmd>, then return
         quit
 partition> print
 Current partition table (original):
 Total disk cylinders available: 1911 + 2 (reserved cylinders)
 
 Part      Tag    Flag     Cylinders        Size            Blocks
   0       root    wm       0 -   61       32.24MB    (62/0/0)     66030
   1       swap    wu      62 -  185       64.48MB    (124/0/0)   132060
   2     backup    wu       0 - 1910      993.76MB    (1911/0/0) 2035215
   3 unassigned    wm       0               0         (0/0/0)          0
   4 unassigned    wm       0               0         (0/0/0)          0
   5 unassigned    wm       0               0         (0/0/0)          0
   6        usr    wm     186 - 1910      897.03MB    (1725/0/0) 1837125
   7 unassigned    wm       0               0         (0/0/0)          0
 
 partition> 3
 Part      Tag    Flag     Cylinders        Size            Blocks
   3 unassigned    wm       0               0         (0/0/0)          0
 
 Enter partition id tag[unassigned]: 
 Enter partition permission flags[wm]: 
 Enter new starting cyl[0]: 186
 Enter partition size[0b, 0c, 0.00mb, 0.00gb]: 200c
 partition> print
 Current partition table (unnamed):
 Total disk cylinders available: 1911 + 2 (reserved cylinders)
 
 Part      Tag    Flag     Cylinders        Size            Blocks
   0       root    wm       0 -   61       32.24MB    (62/0/0)     66030
   1       swap    wu      62 -  185       64.48MB    (124/0/0)   132060
   2     backup    wu       0 - 1910      993.76MB    (1911/0/0) 2035215
   3 unassigned    wm     186 -  385      104.00MB    (200/0/0)   213000
   4 unassigned    wm       0               0         (0/0/0)          0
   5 unassigned    wm       0               0         (0/0/0)          0
   6        usr    wm     186 - 1910      897.03MB    (1725/0/0) 1837125
   7 unassigned    wm       0               0         (0/0/0)          0
 
 partition> 4
 Part      Tag    Flag     Cylinders        Size            Blocks
   4 unassigned    wm       0               0         (0/0/0)          0
 
 Enter partition id tag[unassigned]: 
 Enter partition permission flags[wm]: 
 Enter new starting cyl[0]: 386
 Enter partition size[0b, 0c, 0.00mb, 0.00gb]: 1525c
 partition> print
 Current partition table (unnamed):
 Total disk cylinders available: 1911 + 2 (reserved cylinders)
 
 Part      Tag    Flag     Cylinders        Size            Blocks
   0       root    wm       0 -   61       32.24MB    (62/0/0)     66030
   1       swap    wu      62 -  185       64.48MB    (124/0/0)   132060
   2     backup    wu       0 - 1910      993.76MB    (1911/0/0) 2035215
   3 unassigned    wm     186 -  385      104.00MB    (200/0/0)   213000
   4 unassigned    wm     386 - 1910      793.03MB    (1525/0/0) 1624125
   5 unassigned    wm       0               0         (0/0/0)          0
   6        usr    wm     186 - 1910      897.03MB    (1725/0/0) 1837125
   7 unassigned    wm       0               0         (0/0/0)          0
 
 partition> label
 Ready to label disk, continue? y
 
 partition> quit
 
 
 FORMAT MENU:
         disk       - select a disk
         type       - select (define) a disk type
         partition  - select (define) a partition table
         current    - describe the current disk
         format     - format and analyze the disk
         repair     - repair a defective sector
         label      - write label to the disk
         analyze    - surface analysis
         defect     - defect list management
         backup     - search for backup labels
         verify     - read and display labels
         save       - save new disk/partition definitions
         inquiry    - show vendor, product and revision
         volname    - set 8-character volume name
         !<cmd>     - execute <cmd>, then return
         quit
 format> quit
 #

What's happening:
- Select disk: there are two in the system.
- Go to partition submenu.
- Give starting points and sizes of partitions in cylinders.
- label disk to make partitions permanent.

Partitioning caveats

Be prepared to lose everything on the disk.
Partitions can overlap only if one is unused.
Overlapping partitions in use creates great havoc!
Creating unused holes does nothing except waste space.

Creating a filesystem

Use the newfs command:

 # newfs /dev/rdsk/c0t2d0s3
 newfs: construct a new file system /dev/rdsk/c0t2d0s3: (y/n)? y
 /dev/rdsk/c0t2d0s3:     213000 sectors in 200 cylinders of 15 tracks, 71 sectors
         104.0MB in 13 cyl groups (16 c/g, 8.32MB/g, 4032 i/g)
 super-block backups (for fsck -F ufs -o b=#) at:
  32, 17152, 34272, 51392, 68512, 85632, 102752, 119872, 136992, 154112, 171232,
  188352, 205472,

What's happening:
- Structure disk.
- Add super-blocks of information about the whole disk.

Filesystem creation caveats

Erases everything.
Super-blocks allow recovery from partial disk crash.
Lots of extra options we'll understand later.

Mounting a filesystem

Use the mount command:

 # df -k
 Filesystem            kbytes    used   avail capacity  Mounted on
 /proc                      0       0       0     0%    /proc
 /dev/dsk/c0t3d0s0      83319   27346   47642    37%    /
 /dev/dsk/c0t3d0s6     746982  475574  211650    70%    /usr
 fd                         0       0       0     0%    /dev/fd
 /dev/dsk/c0t3d0s1      41407   19979   17288    54%    /var
 swap                   99900     284   99616     1%    /tmp
 # mount /dev/dsk/c0t2d0s3 /mnt
 # df -k
 Filesystem            kbytes    used   avail capacity  Mounted on
 /proc                      0       0       0     0%    /proc
 /dev/dsk/c0t3d0s0      83319   27346   47642    37%    /
 /dev/dsk/c0t3d0s6     746982  475574  211650    70%    /usr
 fd                         0       0       0     0%    /dev/fd
 /dev/dsk/c0t3d0s1      41407   19979   17288    54%    /var
 swap                   99896     284   99612     1%    /tmp
 /dev/dsk/c0t2d0s3      99723       9   89742     1%    /mnt
 #

What's happening:
- /dev/dsk/c0t2d0s3: the partition to mount.
- /mnt: the mount point at which to make it visible within the filesystem.
- df -k: lists mounted filesystems.

Now let's put something in the filesystem and unmount it:

 # cd /mnt
 # ls
 lost+found
 # touch foo
 # ls
 foo         lost+found
 # cd ..
 # umount /mnt
 umount: /mnt busy
 (I cd'd OUT of this in another shell)
 # umount /mnt
 # ls /mnt
 #

Mounting caveats

cannot mount one device twice.
device must contain a filesystem.
cannot mount two devices on one mount point.
cannot unmount a device into which a process has cd'd. Must cd out and try again.

Maintaining disks

Arranging for mounting at boot time.
Making disks available to other computers.
Mounting disks from remote computers.
Checking for disk problems.
Dump and restore.

Network file system (NFS)

Network File System
Allows mounting of directories from remote hosts.
Sun design.
Monstrous hack.
Frightfully insecure.

Exporting filesystems

Making a filesystem available on the network is called exporting the filesystem.

Solaris sharing information is contained in /etc/dfs/dfstab:

 #     Place share(1M) commands here for automatic execution
 #     on entering init state 3.
 #
 #     Issue the command '/etc/init.d/nfs.server start' to run the NFS
 #     daemon processes and the share commands, after adding the very
 #     first entry to this file.
 #
 #     share [-F fstype] [ -o options] [-d "<text>"] <pathname> [resource]
 #     .e.g,
 #     share  -F nfs  -o rw=engineering  -d "home dirs"  /export/home2
 share -F nfs -o rw=largo /export/1

Linux sharing information is contained in /etc/exports

 /export/mail           @roots(rw,no_root_squash) @sun(rw) @sgi(rw) @linux(rw)
 /export/comp150c       @sun(rw) @sgi(rw) @linux(rw)

@sun, @roots, @linux: netgroups of stations (can also use hostname).

Importing filesystems

Mount the remote filesystem:

 # mount blackhole:/export/1 /mnt
 # ls /mnt
 foo         lost+found
 # df -k /mnt
 Filesystem            kbytes    used   avail capacity  Mounted on
 blackhole:/export/1    99720       8   89736     1%    /mnt
 #

Remote directory must be shared.
blackhole: NFS server.
largo: NFS client.
/export/1: remote directory.
/mnt: local mount point.

Maintaining disks

sync: make disk agree with memory.

fsck: check a disk for errors in structure.

 # fsck /dev/rdsk/c0t2d0s3
 ** /dev/rdsk/c0t2d0s3
 ** Last Mounted on /export/1
 ** Phase 1 - Check Blocks and Sizes
 ** Phase 2 - Check Pathnames
 ** Phase 3 - Check Connectivity
 ** Phase 4 - Check Reference Counts
 ** Phase 5 - Check Cyl groups
 3 files, 9 used, 99714 free (18 frags, 12462 blocks,  0.0% fragmentation)

Only local disks.
Cannot repair all problems.
Asks before deleting anything.

Looking deeper

So far, you've seen what a technical school trainee sees.
We really have no idea what's going on inside the filesystems.
To really understand filesystems, have to know something about the nature of filesystem and process space in UNIX.

Suprising filesystem facts

At any time, a particular file's contents can be distributed over memory and disk.
When a process writes into a file, the new content is buffered in memory until it's convenient to update the disk.
If another process opens the file, content is read from memory if already in memory, instead of from the real disk.
Updating of the actual disk is done by a separate daemon update.
sync means to make the disk image agree with memory. It tells update to 'flush its buffers' to disk.

Why is `fsck` needed?

Any file (or directory) can be stored partially in memory at any time.
If the system crashes, parts of files can be lost.
fsck cleans up after such partial losses, checking that every block on the disk is in an appropriate state.

Naive Model of disk structure

To develop a basic understanding of a UNIX filesystem, please tolerate this oversimplified model of filesystem content.
A disk contains two large arrays: inodes and blocks.
- An inode is a pointer to file or directory content.
- A block is a physical block of storage, 512 bytes.
- These arrays are kept separate on the disk and have fixed sizes.
Each element of each array can be free or used.
- it's free if no one is using it.
- it's used if it's part of a file.
The super-block contains pointers to five things:
- a list of used inodes
- a list of free inodes
- a list of used blocks
- a list of free blocks
- a root directory from which to start adding content.

Newfs

newfs sets up a super-block and several backups, so that
all inodes are in the free inode list.
all blocks are in the free block list.
An inode is simply a pointer to the head of a list of blocks comprising a file or directory. It also contains ownership and mode information.
A directory is a particular kind of file that contains names and other inode numbers represented by the names.

To create a file,

Allocate an inode for it.
Allocate free blocks to the inode.
Copy content into the sequence of blocks.
Map the inode to a name in some directory.
same thing for directories, they're just files that contain name-to-inode mappings.

Really subtle

If the power fails, the process above will be incomplete.
One must check for power failure during each step.
These checks correspond roughly to the phases of fsck!

To read a file:

use your current working directory and its name to determine a directory entry where the inode of the file resides.
Read the sequence of blocks listed within the inode.

To write a file:

Find its inode.
start writing into the sequence of blocks listed for the inode.
If you run out of space, get new blocks from the free block list.
If you stop writing early, return extra blocks to the free block list.

Special considerations

A symbolic link is just a file that contains the path name of another file or directory.
To make:
```
 ln -s target_path linkname
```

Example:

 % ln -s /g/150/motd motd
 % ls -l motd
 lrwxrwxrwx   1 couch    faculty       11 Feb  3 14:16 motd -&gt; /g/150/motd
 % ln -s ../../var/adm admin
 % ls -l admin
 lrwxrwxrwx   1 couch    faculty       13 Feb  3 14:16 admin -&gt; ../../var/adm

Can span filesystems (e.g., jump from one to another).
Slow to read.
A hard link is a reference from a directory to a file that also may live in another directory.
To make:
```
 ln target_path linkname
```

Example:

 % ls -l junk.c
 -rw-r--r--   1 couch    faculty      228 Jan 31 16:00 junk.c
 % ln junk.c junk.cc
 % ls -l junk.c junk.cc
 -rw-r--r--   2 couch    faculty      228 Jan 31 16:00 junk.c
 -rw-r--r--   2 couch    faculty      228 Jan 31 16:00 junk.cc

Link count was 1, now 2.
Indistinguishable from other file.
Cannot span filesystems.
Implementation: two pointers from different directories to same inode.

Hard links are dangerous

Hard links to directories are mighty dangerous (and only root can access them).
The way one finds a filename from an inode is by upward traversal.
If you make a hard link between two directories in the same hierarchy, the kernel will go into an infinite loop in trying to determine the pathname.

E.g.:

 mkdir /foo; cd /foo
 mkdir bar; cd bar
 mkdir cat; cd cat
 ln /foo/bar dog

After this /foo/bar/cat/dog and /foo/bar are the same exact directory.
So in an upward pass from cat to the root, there'll be an infinite loop!
Hence a normal user can't make a hard link to a directory.

lecture in color

/comp/150NET/notes/root.php
downloaded on Nov-23-2009 03:33:36 PM,
was last modified on Feb-17-2004 10:49:41 PM.
All lecture note content is copyright 2004 by
Alva L. Couch, Computer Science, Tufts University

Got root?

So far, you're still normal...

Being root

Becoming root

Stupid root tricks

Ethics

Privacy

Power

Rules of root

Root's view of the world

One example: filesystems

Life of a filesystem

Kinds of filesystems

Investigating filesystems

Studying output of df

File service architecture

Problem of the day #1

Mounting filesystems

Mount rules

True story

Making filesystems permanent

Automounting

Allegro's automount table (assignment 3)

Digging deeper: making a filesystem

Raw disk geometry

Disk concepts:

Partitioning under solaris

Partitioning caveats

Creating a filesystem

Filesystem creation caveats

Mounting a filesystem

Mounting caveats

Maintaining disks

Network file system (NFS)

Exporting filesystems

Importing filesystems

Maintaining disks

Looking deeper

Suprising filesystem facts

Why is fsck needed?

Naive Model of disk structure

Newfs

To create a file,

Really subtle

To read a file:

To write a file:

Special considerations

Hard links are dangerous

Studying output of `df`

Why is `fsck` needed?