Can't browse web pages in Windows 8.1

Many users can ping a web site but can't browse web pages after upgrading to Windows 8.1. Here are two possible reason to solve this issue.
1: DNS problem:
If you can't browse web page through web link(www.google.com) but can browse  through IP address (74.125.226.18), it should be a DNS issue.
Solution:
In Control Panel, open Properties page of Network Adapter, set the DNS as static DNS and set it as 8.8.8.8(Google Public DNS).

2: TCP stack corruption
If you can't browse web page through web link(www.google.com) but can't browse  through IP address (74.125.226.18), it should be a TCP issue.
Solution:
(1) Right click cmd and choose Run as Adminitrator
(2) Run "netsh int ip reset"
(3) Run "netsh winsock reset"

Setup first web application with Django on RedHat

This artical introduces how to setup a simple web application with Django on RedHat, also describes how to troubleshoot the potential errors.

0: Prerequisite Without these component, you may meet errors in the following steps.
rpm -ivh /mntRedHat57/Server/zlib-devel-1.2.3-4.el5.x86_64.rpm
yum install /mntRedHat57/Server/openssl-devel-0.9.8e-20.el5.x86_64.rpm
yum install /mntRedHat57/Server/sqlite-devel-3.3.6-5.x86_64.rpm

1: Install Python 2.7.5
(1) wget http://www.python.org/ftp/python/2.7.5/Python-2.7.5.tgz
(2) cd Python-2.7.5
(3) ./configure
(4) make
(5) export PATH=/root/installer/Python-2.7.5:$PATH
Set Python 2.7.5 as default python, /root/installer/Python-2.7.5 is the Python installation path
(6) Check that Python 2.7.5 can work well

2: Install PIP
(1) Download get-pip.py from this page https://pip.pypa.io/en/latest/installing.html
(2) python get-pip.py

3: Install Django
(1) pip install Django
If you meet "bad interpreter: Permission Denied", go to /usr/local/bin/pip, replace first line #!/usr/local/bin/python2.7 with your actual Python path #!/root/installer/Python-2.7.5/python

Check Django is installed by below command.

python -c "import django; print(django.get_version())"

(2) django-admin.py startproject mysite
Go to a location you want to save web app, then run this command to start a new web app.
(3) python manage.py runserver 8080
Start the web server then you can access 127.0.0.1:8080. The default port number is 8000 if you don't specify. Now the web server can be accessed only in the local host. If you access this web server from another machine, you will meet "Page is not available" or 404 not found error. You need to specify the IP address in the command as below step.
(4) python manage.py runserver IP:8080
With server IP info in the command, the new web server can be accessed from another machine. Below is the expected output.
[root@Ifx mysite]# python ../manage.py runserver 172.19.112.29:8082
Validating models...
0 errors found
May 22, 2014 - 15:15:24
Django version 1.6.5, using settings 'mysite.settings'
Starting development server at http://172.19.112.29:8082/
Quit the server with CONTROL-C.

4:Follow this link to evolution your web app

Perform table restore in Informix

This article introduces how to perform table level restore in IBM Informix with the help of Archecker utility.
Informix Dynamic Server V10.0 (IDS) provides you with the ability to easily extract data in tables of databases from a 0-level backup to a specified point in time. The extracted data can be loaded in internal, external tables or ASCII files. 

1. Archecker utility

The archecker utility uses a configuration file to set certain parameters. Set the AC_CONFIG environment variable to the full pathname of the archecker configuration file. By default, the AC_CONFIG environment variable is set to $INFORMIXDIR/etc/ac_config.std. If you set AC_CONFIG to a user-defined file, you must specify the entire path including the filename. The schema command file is use to specify the source and destination tables and to define the table schema.

The archecker utility uses a schema command file to specify the following:

• Source tablels

• Destination tables

• Table schemas

• Databases

• External tables

• Point in time the table is restored to

2. Table restore

Restore a specific table or set of tables that have previously been backed up with ON-Bar or ontape.

When reading the command file, archecker determines whether to perform a physical restore only or a physical restore followed by a logical restore. By default, archecker performs a physical and logical restore. If you use the WITH NO LOG clause, archecker does not perform a logical restore.

After a physical restore, logical recovery can further restore tables to a user-specified point in time. To do this, the archecker utility reads backed-up logical logs, converts them to SQL statements, and then replays these statements to restore data.

3. Archecker command

>>-archecker---------------------------------------------------->

>--+-+-+- -b-+--| Table-level restore |--+-----+-+--+-----+--+-----+-+-><

   | | '- -t-'                           '- -d-' |  '- -v-'  '- -s-' |   

   | +- -D---------------------------------------+                   |   

   | '- -i---------------------------------------'                   |   

   +- -V-------------------------------------------------------------+   

   '- -version-------------------------------------------------------'   

Table-level restore

|--+------------------------------------------------+-----------|

   '- -X--+---------------------------------------+-'   

          +-+---------------+--+----------------+-+     

          | '- -f--cmd_file-'  '- -l--+-phys--+-' |     

          |                           +-stage-+   |     

          |                           '-apply-'   |     

          '- -D-----------------------------------'     

Element	Description
-b	Provides direct XBSA access for backups created with ON-Bar.
-d	Deletes previous archecker restore files, except the archecker message log. For more information, see When to delete restore files.
-D	Deletes previous archecker restore files, except the archecker message log, and then exits. The -D option can be used with the -X option to delete previous restore files plus any table-level-restore working tables in the sysutilsdatabase. For more information, see When to delete restore files.
-f cmdfile	Specifies that archecker use the command file specified by cmdfile. option overrides the value of the AC_SCHEMA configuration parameter. For more information, see Schema command file.
-i	Manually initializes the system.
-lphys,stage,apply	Specifies the level of logical restore: phys Starts a logical restore of the system, but stops after physical recovery is complete. The backed up logical logs must be available. stage After physical recovery is complete, extracts the logical logs from the storage manager and stages them in their corresponding tables, and starts the stager. apply Starts the applier. The applier takes the transactions stored in the stage tables and converts them to SQL and replays the operations. The default level of logical restore if -l is not listed is -lphys,stage,apply. You can  specify any combination of the logical restore levels, separated with commas. Spaces are not allowed between -l and levels. For more information, see Manually control a logical restore.
-s	Prints a status message to the screen.
-t	Specifies ontape as the backup utility.
-v	Specifies verbose mode.
-X	Specifies a table-level restore.
-V	Displays IBM® Informix® version information.
-version	Displays additional version information about the build operation system, build number,  and build date for IBM Informix.

4. An example to restore a table

Here we extract a table called customer from the most recent backup of dbspace and place the data in table customer.

Create a new command file to cmdfile1.cmd in directory ‘/Informix’ and edit thecommand file as follows:

database stores7;

create table customer

  (

    customer_num serial not null ,

    fname char(15),

    lname char(15),

    company char(20),

    address1 char(20),

    address2 char(20),

    city char(15),

    state char(2),

    zipcode char(5),

    phone char(18),

    primary key (customer_num)

  ) in datadbs;

insert into customer select * from customer;

Restore customer table with the command: archecker -bvs -f /informix/cmdfile1.cmd

Basic GDB

This article intends to help you to master GDB in five minutes. Note that pressing button "ENTER" is to repeat the previous gdb command.

1: Launch GDB
1.1 Type gdb and executable file
-bash-3.2$ gdb executable file
1.2 Type gdb first then attach to the process
-bash-3.2$ gdb
(gdb) attach 21429
Attaching to process 21429
We can get the process id with command ps-ef.
1.3 Launch gdb with core dump file
bash-3.2$ gdb ./readdir core
1.4 Launch gdb with arguments

gdb --args path/to/executable -every -arg you can=think < of

1.5 Launch gdb with commands file

gdb -x commandFile execfile

set auto-load safe-path /
set   solib-search-path /tmp/debug_dir/Base:/tmp/debug_dir/dep_libs:/lib
set   solib-absolute-prefix /tmp/debug_dir/:/tmp/debug_dir/dep_libs:/lib
file  /tmp/debug_dir/Base/cvd 
core  /tmp/debug_dir/core
bt
thread apply all bt

2: Breakpoint management
2.1 Set breakpoint
2.1.1 Break at line number
(gdb) break filename:line_number
2.1.2 Break at a function
(gdb) break function_name
2.1.3 Conditional break which only breaks when some conditions are hit
(gdb) break line-or-function if condition

break x if strcmp(y,"hello") == 0

Num Type Disp Enb Address What 
1 breakpoint keep y 0x000028bc in init_random at qsort2.c:155 
2 breakpoint keep y 0x0000291c in init_organ at qsort2.c:168

2.2 Show breakpoints info
(gdb) info break

2.3 Delete breakpoint
(gdb) delete breakpoint number // Delete one breakpoint whose number is number
(gdb) delete breakpoint  //Delete all breakpoints

2.4 Disable/Enable breakpoint
(gdb) disable breakpoint number
(gdb) enable breakpoint number

2.5 Delete all breakpoints on a certain line of code
(gdb) clean filename:linenumber

3: Watch the context
Watch the context at some breakpoints or crash.
(gdb) bt
(gdb) where

4: Variable management
4.1 Print a variable
(gdb) p variablename
      Print variable in HEX
(gdb) p "%s", stringVariableName
      Print string variable in string format
(gdb) p /x variablename
      Print variable in Binary
(gdb) p /t variablename
4.2 Set a variable to a new value
(gdb) set variablename=value

4.3 Show variable type or strcuture
(gdb) whatis variablename
(gdb) ptype variablename

5: Function management
5.l call functionname
(gdb) call gen_and_sork( 1234,1,0 )
(gdb) call printf(“abcd”)
$1=4
5.2 Finish current executing function and show the return value.
(gdb) finish

6: Ignore an line
(gdb) jump +1
jumps to the next line line i.e. skipping the current line. You may also want to combine it with tbreak +1 to set a temporary breakpoint at the jump target.

7: Run a program with command line arguments
(gdb) gdb --args executibleFileName arg1 arg2...

7: dbx -p (path to exe or libs)

8: Get code address in core file
1) info sharedlibrary: show begin and end address of each .so file
2) x/10i (any virtual address)
3) p/x: calculate addr
4) bt: show virtual addr for each call

10: Important files
ld-linux-x86-64.so.2, /lib64/libdl.so.2 /lib64/librt.so.1 /lib64/libpthread.so.0, libc.so.6(OES), libstdc++.so.6, libthread_db-1.0.so, libthread_db.so.1, libpthread-2.11.3.so, libpthread.so.0

11: Ignore signals

 handle SIGUSR1 noprint nostop

8: Specify symbol files

gdb) symbol-file <path-of-symbol-file>

9: thread debug

set scheduler-locking on/off: Only one thread execute at a time

10: Important files: /lib64/libdl.so.2 /lib64/librt.so.1 /lib64/libpthread.so.0, ld-linux-x86-64.so.2

11: set step-mode on

Determinism in real time system

Determinism measures the consistency of the specified time interval between events. Determinism is essential to the real time application. If the event doesn't happen in time as expected, the consequence of the real time application is extreme serious, for example the air bag doesn't unfold in time in a car accident. This article introduces the conceptions of jitter and determinism, then describes what needs to be considered when we implement a real time application.

1: Jitter and determinism

Jitter is the difference between the expected and the actual timing responses. Determinism measures the consistency of the specified time interval between events.

2: Characters for real time operating system

To achieve high determinism, real time operating system supports below features

1): Minimize feature set

Real time operating system focuses on necessary features, such as thread scheduling and file system management. It deletes all unnecessary or resource consumption features, such as GUI and security module.

2): Restricted priority thread scheduler

Thread scheduling based on thread priority. Perform preemptive scheduling among threads of different priority, and perform round-robin scheduling among threads of equal priority.

3): Minimize interrupt and thread switching latencies

3: Design guideline for real time application

3.1 Memory allocation

3.1.1 Steady state allocation

The component should have no memory allocations or de-allocations in the steady state, even in non-time-critical threads. Steady state is defined as the stable operating state of the component. Allocations and de-allocations may occur during initialization and shutdown, which must be processes with a defined beginning and end, which happen at times predictable to the user.

Thrashing the memory manager can cause memory to become fragmented, which will decay performance of the allocator over time. This applies to explicit calls to malloc/free and new/delete as well as “hidden” calls such as via STL containers.  

3.1.2 STL

While the use of STL is encouraged in C++ code, it is important to be careful about how STL implementations handle memory management. In many cases, STL provides no guarantees about which operations may cause an “under-the-hood” allocation. In general, if an operation is said to “invalidate iterators,” then it can be assumed that this operation is allowed to cause an allocation under the hood. But the reverse is not true: some operations do not invalidate iterators, but do cause allocations, such as std::list::push_back().

The only completely safe assumption is that anything that adds or removes elements from an STL container may cause a memory management operation. One way to mitigate this problem is to use the “pool” feature of the Boost libraries. This provides a pooled allocator that complies with the STL allocator interface. Using this, you are guaranteed that no memory operations will take place at the OS level unless the container grows beyond the maximum size that it has ever been. However, keep in mind that the pool still uses a mutex, which can lead to “priority inversion” (see the Threading section below).

There are some exceptional cases where certain STL containers provide memory management facilities. The notable example is the std::vector::reserve() method. Using this method, you are guaranteed never to hit the memory allocator unless the container exceeds the reserved size. Very few STL containers provide similar facilities however, making vector one of the few completely safe containers from a memory management standpoint.

3.2 Threading

3.2.1 Priority inversion

Priority inversion can happen any time a high-priority thread depends on a low-priority thread to make progress. The simplest example is when a high-priority thread and a low-priority thread compete for the same mutex, and a medium-priority thread preempts the low-priority one. Since the high-priority thread cannot wake up before the low-priority thread lets go of the mutex, and the low-priority thread cannot wake up until the medium-priority thread yields, the high- and medium-priority threads have inverted priority.

Most RTOSes (including the ones we use: VxWorks and Phar Lap) counter priority inversion with a technique called “priority inheritance.” In this protocol, any thread that owns a mutex “inherits” the priority of the highest-priority thread currently waiting for that mutex. In the above example, that would elevate the low-priority thread above the medium-priority thread long enough to get its work done and let the high-priority thread run. However, it is important to note that this does not eliminate priority inversion; it only establishes a provable worst-case bound. That bound is the summation of the time that all mutexes shared by threads of differing priority are held. If you hold mutexes for a long time, this can still be very bad.

In order to minimize mutex-based priority inversion, observe the following guidelines:

• Mutexes should be held for as little time as possible. Note that this is a trade-off: overly fine-grained mutexes can hurt system performance from repeated acquiring/releasing.
• To combat the overhead of fine-grained mutexes, eliminate the need for mutexes where possible. A common technique is to create copies of shared data that each thread owns and can operate on independently. Then, when it is time to do something with that data, you can grab a mutex for a short amount of time and “apply” it (by writing it to hardware for instance). Note this is not the same as “thread-local storage,” which is a facility provided by the OS directly. TLS implementations vary widely from OS to OS, and its use is not recommended in portable software.
• If it is possible to eliminate the mutex by using atomic operations (such as atomic increment/decrement) or wait-free data structures, then do so. Unfortunately no standard implementations of wait-free queues and so forth are available.

3.2.2 Event-based priority inversion

Event-based priority inversion occurs when a high-priority thread waits on an event or signal from a low-priority thread, and that low-priority thread becomes preempted and cannot progress. Priority inversion from events or signals is often worse than mutex-based priority inversion, since OSes typically have no built-in countermeasure for it. Unlike mutexes, which have a concept of “ownership,” no one owns an event, so the OS cannot tell whose priority to elevate.

A common example is a high priority thread which consumes data coming in off the network. The network is serviced by a low priority thread that sets an event when traffic comes in. The high priority thread waits on that event. If the low-priority thread becomes preempted, it will stall the high-priority thread.

The simple guideline is do not design systems that do this. High priority threads must be able to progress without depending on progress from lower-priority threads in a real-time system.

3.3.3 Round robin

Most RTOSes implement a scheme known as “round robin” for multiple threads at the same priority. Under this scheme, a thread is allowed to run for a certain period of time (usually a handful of milliseconds), then it is swapped out with the next thread in line at the same priority, and so on in a circular fashion. This leads to the decision to run several “unimportant” threads at the same priority in a system, under the assumption that they will get roughly equal time. However, this practice is bad for 2 reasons:

• Thread context switch overhead is usually significant, even at a round robin period of 10 msec.
• More importantly, the assumption of equal time for round robin threads is incorrect! If a thread yields its time slice prematurely for ANY reason (mutex acquire, sleep, explicit yield), then its time slice is gone until its turn comes around again. For example, say you have threads A, B, and C. A and B compete for the same mutex, and they acquire and release it every 10 microseconds. C runs wide open. What will happen is A will run for 10 microseconds and go to sleep on the mutex. B will do the same. Then C gets a full time slice for 10 milliseconds. A and B get 1/1000^th the time of C!!!

The second point above can lead to a “quasi-starvation” condition if a thread yields very frequently (and remember, a yield can mean a simple mutex acquire) compared to the round robin period of the system and other threads at the same priority do not. You should be able to demonstrate that threads scheduled at identical priority to other threads will not undergo quasi-starvation.

The guideline for this situation is this: if you have a thread in your driver that does any significant amount of work, it should run at a unique priority, or at least only share a priority with other threads in your driver that you can control. Either it’s a background thread that runs lower than everything else, or you knowingly allow it to preempt everything below it. Simply relying on the RTOS scheduler to “do the right thing” almost always produces unwanted results.

3.2.3 Deadlocks

Deadlocks occur when 2 or more threads are waiting on each other to do something, and neither can progress until that happens. It’s difficult to address this situation with a coding convention, apart from “always acquire and release locks in consistent LIFO order.” I am including it here because the working group has reported that deadlocks are the most prevalent problem in RT software.

A common cause of deadlocks is the “hidden mutex.” Your component uses a library that acquires a mutex internally. Also, this library issues callbacks to your component, and the same mutex is owned by the library when the callback is issued (this is a lot more common than you’d hope). Of course, if your component owns a mutex when it calls into the library, then also tries to get the same mutex from within a callback issued by the library, you have a deadlock. The only solutions to this are:

• Contact the developer of the library for details on mutexes owned when callbacks are issued (not always possible).
• Never acquire a mutex in your callback that your component owns when calling into the library. This may require you to perform actual processing in response to a callback from an internal thread to your driver.

A bug in Informix V12 installation in Linux?

Informix V12 installation reports that "User informix does not exist". But when I try to manually add user informix, it says user informix exists. This issue happened in two different Linux machines.

The solution is simple. Trigger the installation again then error disappears. I guess that the installation procedure checks user informix existence before creation during to race condition issue.


1: Errors in Console:
Informix User/Group Creation Error
An error has occured during the creation of the specified user or group. See the log for more detail.
PRESS <ENTER> TO ACCEPT THE FOLLOWING (OK):

2: Could not add user informix:
[root@Ifx installer]# user informix
-bash: user: command not found
[root@Ifx installer]# groupadd -g 777 informix
groupadd: group informix exists
[root@Ifx installer]# addgroup -g 777 informix
-bash: addgroup: command not found
[root@Ifx installer]# useradd -u 777 -g informix informix
useradd: user informix exists


3: Trigger the installation again then user creation error disappears
[root@Ifx installer]# ./ids_install