Notes for Week 3

1. If you have not yet created a non-root user, do so at this time using the commands
   • useradd -m name, to create the account, and
   • passwd name to give the account a password (where "name" denotes the login name).

     These operations are performed using the "Users and Groups" control panel in Windows.

   Logout as root and log back in as that user, and do most of your work as that user. When you need to work as root, execute the command su (which will prompt you for the root password) to temporarily become root. When you have finished the work which required root privilege, use the exit command to return to being a non-root user.

   There are some nice screen savers that will run under non-root users. Unfortunately, in the past, some of the 3D ones could lock your system up tightly enough to force you to power down. The safest thing is to select the "Blank Screen Only" mode in xscreensaver-command -prefs.

   A systems administrator NEVER walks away from her/his console without first locking the screen.

   You can do this with xscreensaver-command -lock.

   Note that in order to use the 3-D screensavers, you must be a member of the video group. We will learn how to do this in week 6.
2. Examine the file /usr/doc/linux-3.13/devices.txt. It contains a list of all of the hardware device filenames which can appear in /dev. This is not the same as all of the devices supported by Linux. That list is much larger (since many types of devices can accessed by the same device file, and some devices, such as many network devices, do not have /dev entries), and is constantly changing.

   To find out if a device is supported, you have to know which version of the kernel you are running, and what capabilities were built into it. Suppose you wanted to see if your kernel supports the Realtek 8169 Gigabit Ethernet adaptor. You would first examine the

   Kconfigs files in the linux source tree (here collected into one file for your convenience)

   for your Linux distribution to identify the "config" parameter associated with that device. Loading the file (either into your browser, emacs or less), and searching for

   Realtek 8169

   tells you that the configuration parameter is called "R8169". Then examine the configuration file for your kernel:

   /boot/config-3.13.3 was used for our kernel

   (either by loading the file and searching for R8169, or by using the "grep" command:

   grep /boot/config-3.13.3 -e R8169

   grep is like the Windows FIND command after it has been bitten by a radioactive spider.

   grep is your friend. For example, you can use it as a sort of dynamic glossary. Suppose you save a copy of each page of these notes in the directory "notes". To find which pages contain references to "passwd", you would simply

   grep -ile 'passwd' notes/*.html

   which results in the following output:

   notes/week1.html
   notes/week3.html
   notes/week6.html
   notes/weekd.html
   notes/weeke.html

   This tells you that the passwd command (or something related to it) is discussed in the notes for weeks 1, 3, 6, 13 and 14.

   The "i" option tells grep to be "case-insensitive" when searching for the string "passwd"; this allows for matching of the string at, for instance, the beginning of a sentence. The "l" option tells grep to simply "list" the names of the files in which the string occurs, and the "e" option specifies that the next string is the "expression" you which to search for. The asterisk is a "wild-card", which matches any string; here, as often, it is used to specify any filename ending in ".html". We used the asterisk in step 5 of week 2 to copy everything in / to the hard drive.

   Either way, you find the line "CONFIG_R8169=m", which tells you that support for that device has been built into your kernel as a module which can be loaded and unloaded dynamically as needed. If the result had been "CONFIG_R8169=y", the driver was built into the kernel as it is loaded at boot time, and is always in memory, and if the result had been "CONFIG_R8169=n", there would be no support for this device in your kernel.

   Note that newer devices which use the same chip set as an older device will often work with the older device driver.

   Linux support for newer devices, particularly video cards, often lags behind Windows support. When a manufacturer designs a new device, they write a Windows device driver for it and provide both the hardware and software to Microsoft so that it can be integrated into the operating system. When someone in the Linux community wants support for a new device, they have to buy one of the devices for themselves, either write a new driver or modify an existing one to work with it, and then give it to the open source community for testing and integration.
3. Section 4 of the man pages contains descriptions of many of the device files. Of particular interest are:
   • character devices with major device number
     • 1 (memory device files; see the mem, null, zero and random man pages) and
     • 4 (virtual console and serial port device files; see the tty man page);
   • and block devices with major number
     • 8 (SCSI device files; see the sd man page) and
     • 11 (SCSI CD-ROM device files).

   Suppose you wanted to find out quickly what the major and minor device numbers were for SCSI drives sdb1 and sdb2. Here are several ways to do this:

   • ls -l /dev/sdb[12]
   • ls -l /dev/sdb{1,2}
   • grep /usr/doc/linux-3.13/devices.txt -e '/dev/sdb' -C10

   The first two provide a "long" list of the devices files, which show the numbers preceding the creation date, as well as the device type (block) in the first letter of the output. The "[12]" is a way of allowing either character to match in the filename (and can be used with grep), and the "{1,2}" is similar but can be expanded to include whole strings in the place of the numerals.

   These two methods only work if the device file exists. More about this when we discuss udevd next week.

   The last way illustrates both a way to see the 10 lines surrounding any grep match (the "Context"), and that sometimes you get much more information than you expected!
4. File management under Linux is handled by the Virtual File System. Since Linux supports many different types of filesystems (Ext2, FAT, VFAT, ISO-9660, HPFS, NFS, NCP, NTFS, etc.), each filesystem type has its own driver and the VFS serves as a common interface between them and the file access routines that handle opening, reading, writing, closing, etc., of files and directories. So, for instance, an ls command will produce compatible results no matter which filesystem is being referenced. The VFS maintains a common inode and directory cache for all of the filesystems, and the filesystem drivers use the same disk cache and drivers. All currently accessible files and directories have path names relative to the same directory tree root, called "/".

   Examine /boot/config-3.13.3 and Kconfigs to see which filesystem types are supported by this version of Linux.

   In general, Windows only supports three types of filesystems: FAT, NTFS, and ISO-9660.

5. Some useful file and filesystem commands:
   • ls directory is used to list the contents of a directory (try -l and -a options)

     A folder is an icon representing a directory; if the directory is not specified, the current working directory is listed. The complete (or absolute) path to a directory starts with the root directory, such as "/etc/rc.d/init.d". By convention, filenames beginning with a period are not listed unless the "a" option is used to list "all" filenames. A single period denotes the current directory, and two periods denotes the parent directory.

     (This is sort of like the DIR command in Windows.)
   • There are many ways to display all or part of the contents of a file:
     • cat file displays the whole file (like TYPE in Windows);
     • less file displays the file a page at a time (like MORE in Windows, but more powerful);
     • head -nnumber file displays the first number lines of a file;
     • tail -nnumber file displays the last number lines of a file;

     and of course, you can always use emacs or nano (if X-Windows is not running), or another editor, to examine a file.

   • cd directory changes the current working directory (same in Windows)
   • mkdir directory creates a new directory (same in Windows)
   • rmdir directory removes a directory (must be empty) (same in Windows)
   • mv oldfile newfile moves a file from one location (or name) to another (like MOVE or RENAME in Windows)
   • ln -s file linkname creates a symbolic link to a file or directory (like a shortcut in Windows)
   • cp file filecopy copies a file (-a can be used to copy directories; if a directory is specified instead of "filecopy", a copy of the file with the same name is created in the directory (like COPY and XCOPY in Windows)
   • rm file removes (deletes) a file (like DEL in Windows)

     rm -r directory will delete all of the files and subdirectories in directory, and then rmdir the directory itself.

   • Use df to display space utilization on filesystems. You can do this in Windows by opening "My Computer", right-clicking on the drive and choosing Properties.
   • To see specific information about the space taken up by files in specific directories, use (for example) du -s /usr/*.

     The du command can be invaluable in finding out which directories in your system are taking up the most space.

   • Software hangs can often arise from file locks or conflicts; the command

     fuser -v file

     displays information about processes which have a given file or directory open. As is the case for many commands, the "v" option tells fuser to be "verbose".
   • hdparm can be used to interrogate and modify drive settings. In the lab, hdparm -iI /dev/sdb will display information about the hard drive (the first part may not be completely current; the second part is).

     The command hdparm -tT /dev/sdb will run timing tests on it; these are an indicator of potential throughput.

     ANY OTHER USAGE OF HDPARM IS DANGEROUS; if there is something you want to try in the lab, check first with the instructor. If you damage the hard drive using hdparm without having first obtained permission from the instructor, you will be subject to the withdrawal/expulsion policy outlined in the syllabus under the section "Tests and Grades".

   • The stat command can be used to view the details of a file's inode.
   • The sync command forces synchronization of the mounted filesystems with the VFS and device cache. Periodically executing the sync command ensures that no data will be lost in the event of a power failure.

     Note that some hard drives also contain cache, which is not guaranteed to be flushed with a sync command.

   • The touch command updates a file's modification time to the current time without changing any other attributes of the file (if the file does not yet exist, touch will create an inode for it, but no space will be assigned for its contents). Hence

     touch file

     creates a sort of filesystem time stamp which can be used to flag important events in the life of a filesystem. For instance, if you touch /root/.installed immediately after installing your OS, find / -xdev -newer /root/.installed will find all files and directories modified since installation.

     Note that the find command is not the same as the FIND command in Windows. We will discuss the find command in more detail in week 8.

6. A flash drive can be mounted using the command mount /media/usbstg. If you use a second flash drive simultaneously, in our lab the additional device address will be sdd1. For sdd1, you will have to mount it explicitly with mount -t vfat /dev/sdd1 /mnt (only root can do this, since the mount points are not listed in /etc/fstab).

   Remember to unmount the drive before removing it physically from the interface; like floppies, there is no guarantee that the filesystem on the drive is in a consistent state if you remove it while it is still mounted.

7. The installation we performed last week left us with two ext2 filesystems: one for the root filesystem, which is mounted read-only, and one for /var, which is writable. Occasionally we will need to make modifications to the root filesystem. For instance, you may wish to create a directory /media/flash as a mount point for a flash drive. To modify the root filesystem, you can re-mount it writable using the command mount -wo remount /. After making the desired modifications, you can then re-mount it read-only using the command mount -ro remount /.
8. An entire filesystem can be copied to disk using the dd command. For example, dd if=/dev/sdc1 of=flash.image will copy the entire first flash drive partition to the file flash.image. This disk image can then be mounted using the loopback interface, and treated like any other filesystem: mount -o loop flash.image /mnt.

   You may have to modprobe loop to load the loopback module if it has not already been loaded.

   We did this in step 5 last week.

   Window's DISKCOPY command is a very limited form of dd.

---

9. EXERCISES for Week 3:
   1. Are there any "hidden" files in the root directory? In /etc? In /dev?
   2. Search all the files in your home directory (but not the subdirectories!) for the string xscreensaver. Display the contents of every file containing that string.
   3. Make a bin directory in your home directory. Copy the files you just found into it. Change to your new bin directory and rename the files there to remove any periods from the names. Delete those files.
   4. What are the device type, major and minor numbers for the device holding /var while running from the DVD?
   5. What are the device type, major and minor numbers for the first loopback device?

©2015, Kenneth R. Koehler. All Rights Reserved. This document may be freely reproduced provided that this copyright notice is included.

Please send comments or suggestions to the author.