This pages contains some of my UNIX tips that you may find it useful and interesting.
 

Command tips
Followings are some of my favour commands to use.

How to find out all the files containing a particular string under a particular directory ?
    # find <directory> -print -exec grep <string> "{}" \;
 

How to remove those Ctrl-H characters output by the man pages, so that you can vi the output of man command ?
Try the following script and name it as rmctlh :

     (ctrlh=`echo '\010'` ; sed "s/.$ctrlh//g")

     Change mode of rmctlh to be user executable, then use it as a filter :

    e.g. $ man man | rmctlh
 

If I just want to remove some of the files under a particular directory, how can I have Unix prompt me to select which one to delete ?
Try : $ find <directory> -ok rm "{}" \;
 

How can I kill all the process with a particular name ?
By making use of the pipeline and some filter features of Unix, try :

     $ ps -ef | grep <process name> | awk '{print $2}' | xargs kill

System administration

How to find out all the same hard links of a particular file ?
   First is to find out the inode # by :

        # ls -i

        Then use following to find out all the files with that inode # :

        # ncheck <inode-#>
 

Which file I should examine if setld fail ?
    Try checking /var/adm/smlogs/setld.log which may have information that you will be interested in.

How to reset the X server ?
Add the -terminate option in the /var/X11/Xserver.conf , then kill the Xserver process (/usr/bin/X11/X). The xdm will automatically restart the Xserver. The Xserver will also be reset each time when you logout after adding the above option.

Unix networking

How to find out the IP address of your host ?
   First find out the interface name by :

        # netstat -i

        Then use following to find out the I.P. address of that interface

        # ifconfig <interface>

        In Solarise, ifconfig provide a -a option which will show all the interface information.
 

Shell programming

How to mirror two directories ?
   Here is a script that I used :
 

C programming

How to find out which library contains a particular symbol ?
  Here is a simple script I use.
 

Performance Tuning
What happen if excessive page in but few page out in the output of vmstat ?
Then most likely, you have set the UBC-borrowprecent too low, thus causing the UBC pages easily been claimed by the VM, since the UBC pages usually are unmodified, thus why not much page out. But if they are accessed again (since hey have been kicked out easily by the VM), then they will be loaded back again to physical memory, causing excessive page in.

What can I do in a low memory Xserver system if memory shortage happen ?
Try to reset the Xserver as the space allocated to Xserver will not be freed until the Xserver is reset.
 

Unix internal & advance topics

If you find the following questions interesting, I recomment the Unix Internal course which will provide more detail information to satisfy your need.

What makes traditional Unix not a real time system ?
There are two factors that make traditional Unix not a real-time operating system :

• Scheduling policy : as the original design of the Unix is to facilitate the interactive processing as in a time-sharing environment. So as to make every user process has a chance to acquire the CPU time, the priority of the process is automatically adjusted when it is executed by the CPU. The longer the process that have been executed by the CPU in its time slice, the lower the priority it will become, so that whenever another process (that usually waiting for I/O) is ready to be executed, e.g. wake-up by the terminal input, then the new process will have higher priorty enough to be executed immediately. This prevents any process from occupying the CPU forever, but it also hinder the real-time applications, as real-time applications may need to be executed before any other processes having chance to run. In fact, this restriction still occurs in 4.4 BSD. Thats why Posix Real-time call are introduced in some commercial Unix system.
• Kernel pre-emption : thats another critical design restriction of traditional Unix. As in Unix system, the kernel will only check whether it needs to do context switch whenever before a running process return from kernel mode to user mode. It means that it'll never preempt a running process when it is in kernel mode such as when a process is in a system call even if another higher priorty process is ready to run (e.g. an interrupt handler has already wake-up another higher priorty process during the execution of the current process's system call). As there is no gauarantee on the upper bound of the time being used in kernel mode, i.e. no gaurantee on response time of any process even in the highest priorty, thats why its not real-time. Again, 4.4 BSD also have this restriction. To solve this problem, some of the kernel code must be re-designed.

What is the difference between BSD & System V Unix?
In the history of Unix, around the late 70's. Unix branches into two different major release, one is the System V release developed by AT&T. While the other is the Berkeley Software Distributions (BSD) version. The BSD project was later then to provide support for the DARPA Internet networking procotols TCP/IP, thats why BSD socket was developed. While the System V implementation provided the STREAMS, an IPC (inter-process communication) mechanism. The internal memory management structure of System V was also different from the Berkeley paging system. Besides the above major difference. The signal handling and terminal session handling are also different between BSD and System V.

Why we need multi-threading in Modern Unix ?

Traditionally to support multi-programming in Unix, usually subprocesses are used. Usually the parent process will fork different child process to handle different things that need to be executed in parallel in a particular task. A typical example is the internet daemon inetd which will fork different child process to handle different internet connections e.g. one telnetd for each telnet session and one ftpd for each ftp session. But whats wrong with this approach. Since to create a sub-process is not an economic job, its quite expressive for using system resources. e.g. allocation of process structure and page tables in a paging system (though most modern Unix uses absolutely lazy evaluation to allocate memory i.e. defer the allocation of physical memory until the last moment, still it needs to allocate stack etc.). Thus threads are introduced in modern Unix. Threads are usually refered as light weight objects and relatively less overhead to create. Since all the threads in a process shares the whole virtual addressing space of that process, only different stacks (and a seperate thread structure in the kernel for kernel thread) are need to be allocated. Its more efficient and less expensive.

What is the WCHAN field in the output of ps -el ?
WCHAN is called the wait channel. When a process is going to wait for a system resource, it'll be assigned to a wait channel, usually it'll be an address related to the resource, and later the wakeup call will be passed with this address to wait the process up. e.g. for reading the hard-disk, the WCHAN will be the address of the buffer that going to receive the data. Then once data is arrived, the interrupt handler will use the same address to wakeup the waiting process.

vi = ex ?
Try the following if you have super user privilage.

     # cd /usr/bin
     # rm vi
     # mv ex vi
     # vi

See, vi is exactly equal to ex ! In fact, if you use the ls -i to check their i-node number, they are the same. i.e. it means that they are hard-link to each other. But how can they be different during execution and what's the advantage of doing so ? Thats the advantage of using the first argument (in fact the zero-th) in a C program. As in every C program, the zero-th argument is the name of the file itself, by checking this argument, the program can branch to different starting point. But why the original developer didn't seperate them into two different programs ? The answer is for the traditional Unix, share library is not supported. As vi and ex are very popular programs that likely to be executed by different users in the system at the same time. If they are the same executable, the kernel will share the common text portion (i.e. executable part of the program file) with all the users, in order to save physical memory. In fact, many Unix programs use this tricks. Verify with passwd and chsh program.
 

What is the difference of symmetric multi-processing (SMP) and asymmetic multi-processing ?
To know more, the Unix Internal course will be interested to you.

What are kernel threads and user threads ?
See the diagram.

When will be the running process go to the kernel mode ?
What are those defunct process shown in the output of ps ?
What are the limitations of the traditional Unix file system ?
In the traditional design of both s5fs (System V file system) and Berkeley's ffs (Fast Filesystem), the inode table can only reside in only one disk (or a partition of the disk) and the index of the inode table which points to the actual physical data block can also only address the local disk (or partition). Thus files cannot split over different physical disks (or partitions). Moreover Unix are usually designed for 32 bits machines (except those early 16-bits dinosaur machines!), thus the offset of the files passed into most of the filesystem related system calls e.g. read, write are usually an integer. This imposes another logical maximun for the file size to be 31 bits (most significant bit as sign bit). Moreover, since the inode table is something like an on disk array design, so you need to estimate how large of the table when you first initalize your file system by specific the inode density in the mkfs command. Once the table is full, you don't have any way to increase it -- only way is to backup all your files, re-initialize the filesystem and restore all your data -- how clumsy it is ! To address the above problem, Digital Unix introduce the Advance File System, try attend the Unix Internel course for more detail.

What are the restrictions of traditonal buffer cache design ?

Perl & CGI

How to use Perl to emulate the UNIX wall command ?
   Click here for the example.
 

Back to my home page