Cyber Note

Android Git server always fatel error?

2010-02-13T15:49:00.005+08:00

WTF..Android repo init -u always return the error:

fatal: Unable to look up android.git.kernel.org (port 9418) (Name or service not known)

Someone said that this may be caused by the DNS resolving issue.
So I did some work on my /etc/hosts as adding a line:

149.20.20.141 android.git.kernel.org

It depends on your network environment and ISP...

OK, now repo tool finally works!! What an idiot script it is!!
...Orz

Chromium OS on TOSHIBA NB300

2010-01-14T15:23:00.005+08:00

Recently I am working on the Google Chromium OS for my TOSHIBA netbook.
I've successfully build the OS image and create a live USB pen drive.
For convenience, I install it into HDD...;-)

But, Chromium OS is useless if you don't have an internet connection........

Some notes here:
1. root file system is read only, so need to remount to read-write mode.
$ sudo mount -o remount,rw /

2. My Chromium OS kernel is 2.6.30-chromeos-intel-menlow.
In this archived version of Chromium OS, I setup some Chrome browser default settings in this file: /usr/bin/chromeos-chrome (This is a shell script text executable).

You can setup your own start-up pages and tabs here within this script.
Or you setup additional property values to Chrome browser.

Here I set my favorite proxy server (maybe you know Chromium OS got some problems for internet proxy setting........).

Use vi to edit it (remount root fs to rw first), find the last while-do section that including a variable called "$CHROME".

Add one string to the top of this variable such as:
"$CHROME" --proxy-server=aaa.bbb.ccc.ddd:8080 \
...
[...rest of the code..]
...

Else, you can just read the script for fun, it's interesting...;-)

My next step is connecting my Chromium OS with wifi.
And maybe I will tune the ACPI subsystem by some extra programming (If I got free time to do this..).

Ethernet controller performance is perfect (Realtek 8101E/8102E, PCI-E)
My Atheros wifi card is well detected (AR9285, PCI-E) and kernel module works fine (ath9k, mac80211 and cfg80211).

NPNT...X-D
For fun, I also change the login screen background image..

Google v.s China? BYE BYE~~

2010-01-13T16:39:00.001+08:00

Official Google Blog: A new approach to China

FUCKING YEAH, Google ROCKS!!

Standard Linux uvcvideo doesn't work for Bison camera module 5986:0241

2009-10-28T16:10:00.002+08:00

Usually, current uvcvideo driver kernel module can handle all uvc complaint devices.
But I got a special module which ID is 5986:0241, can not work w/ it.
You can see a dmesg log as "No valid video chain found.", and driver never drive this device.
There is no device file created under /dev and /sys.

This is a FW bug, so that driver doesn't parse video steaming data well.
I patch the uvcvideo source to skip this checking.....:-(
Not a good way but it works.

Use kernel 2.6.31.5 source as an example to patch the driver:

-> drivers/media/video/uvc/uvc_driver.c, Line 430:
modify "return -EINVAL;" to "break;"

-> drivers/media/video/uvc/uvc_driver.c, Line 723:
modify "goto error;" to "break;"

Save and compile the uvcvideo.ko again.
After install this new kernel module, your 5986:0241 camera should be back to work.

RHEL / CentOS 5.3 never read MAC address on SiS 190/191 NIC

2009-09-02T11:26:00.006+08:00

Oops........
This issue was reported for a while from my FAE.
The kernel of default RHEL 5.3 is v2.6.18........(WTF..)

I personally think that it's a bug of sis190.c driver but I don't have time to dig into the source code.
The NIC MAC address can not be read correctly from EEPROM.

I use sis190.c and mii.c from kernel v2.6.19.7 w/o any modification, and then compile them on RHEL 5.3.
Well, they still taint kernel a bit because no version magic.
But works well (so far..), and my MAC address of SiS 190/191 NIC was now detected.

Use root permission and just # modprobe -rfv sis190
Make mii.ko and sis190.ko by yourself. If you see warnings, forget about it now......
# install -m 644 *.ko /lib/modules/`uname -r`/kernel/drivers/net/
You can make a backup of original mii.ko and sis190.ko if you want.
# depmod -ae
# modprobe -fv sis190

Linux SiS 190/191 NIC lose traffic on 64bit kernel with 4GB RAM

2009-07-22T11:31:00.003+08:00

SiS190.x driver got a bug on 64bit Linux kernel with 4GB RAM.
NIC can get IP via DHCP correctly but can not have any traffic any more...Orz

If you limit RAM size while bootup to "mem=3G", everything works fine ( Of course......XD ).

See this for bugzilla report: here.

I test the patch on Mandriva 2009.1 and it works.
But I never test it on Ubuntu or any other distro yet.
Here is the duplicated issue on launchpad for Ubuntu.

CM Linux is a small distro which is based and modified from Mandriva Linux......囧rz

clevo_laptop v1.0.0 is released

2009-07-06T21:14:00.004+08:00

Finally, I uploaded the first release of Clevo ACPI WMI Linux kernel driver module to Sourceforge.net.
The release is clevo_laptop_1.0.0.tar.gz.

After a long time investigating of BIOS ASL code, I found so many bugs and NGs on their laptops.
For the ACPI supportinf about Linux.
Even I have told them, the BIOS engineers, no body gives a shit about my opinion........

Will keep developing this kernel module to support Linux on Clevo platform as much as possible.
('Cauz BIOS guys don't want to pay efforts to Linux, this is not listed in their spec........)

R.I.P - King of Pop - Michael Jackson 1958-2009

2009-06-26T17:02:00.004+08:00

OMG, this morning I got a bad news that a most widely famous artist - Michael Jackson just past away!

I got almost all his albums that I can bought here.
Since Off the wall, Thriller, Bad, Dangerous, History, Scream, Invincible...etc.
Of course, too many songs he performed can be classics..

Q__Q..

I will miss you, Michael.
I will miss your dancing.
I will miss the performance you made with Slash.

You are always the best!

SourceForge clevo-extra project kick-off

2009-06-16T09:59:00.003+08:00

It is a milestone for myself...;-)

Project UNIX name : clevo-extra
Registered : 2009-06-11 10:17

http://sourceforge.net/projects/clevo-extra

Compiling SiS X driver on Mandriva 2009.1

2009-06-02T13:56:00.007+08:00

Based on this one, I also try to compile and install SiS X driver on Mandriva 2009.1.
Of course I have to install some additional packages.
Use urpmi and all related sw automatically:

Xorg X server is v1.6.1.

1. binutils;
2. make;
3. libxorg-x11-devel;
4. xorg-x11 (this is a meta package);
5. xorg-x11-xvfb (I think this one is optional..);

BTW, I try it on a small distro called Caixa Magica 12.
Can compile and install 'cause it's a Mandriva-based distro.

Compiling SiS X driver on Ubuntu 9.04

2009-05-27T13:59:00.003+08:00

I got new verison of SiS Linux X driver which fixed some bugs and the AUO panel issue.
And again, I have to verify it for our company's product on Ubuntu 9.04 first.
(Company decision........)

Xorg X server is v1.6.0 on Ubuntu 9.04.

Well, some packages I need to install to compile the SiS X driver source code:
1. xserver-xorg-dev;
2. x11proto-*-dev; (Lazy, so I install them all......)
3. libdrm-dev;
4. libgl1-mesa-dev;
5. build-essential; (This is ur call..)

And, of course, ./configure; make, then you can find the driver sis_drv.so under ./src/.libs/.
Nothing special.

Next I have to verify it on Mandriva 2009.1 also.

Inside the Linux reboot.c of kernel 2.6.25

2009-05-20T12:37:00.011+08:00

I got a bug report from one of my customer, they said system hang up after issuing reboot command if using T3400 CPU...= =|||

I can duplicate this issue and I think it is a BIOS bug on our product, because I saw this issue before BIOS POST.

After confirm with our BIOS team members and we use 80 port debug card to see what is going on. We know that BIOS team need to modify their watchdog to monitor this kind of problem......They need to know some information to catch for the system reboot. In Windows, they can monitor it because system will write 0xfe to 0x64. I believe that Linux will do something smiliar to this but I can not figure it out in kernel 2.6.24.7 source code under arch/x86/kernel/reboot_*.c...

Well, starting from Linux kernel 2.6.25, there is only ONE reboot.c under arch/x86/kernel, and they do change some points in the code.

In arch/x86/kernel/reboot.c and line 367 we can see the code:
outb(0xfe, 0x64); /* pulse reset low */

This function is defined in arch/x86/boot/boot.h.

囧rz..
This means that system will do basic port I/O to write to port 0x64 with value 0xfe when system is going to reboot. Bingo! So our BIOS team will be very happy with it........'cause they now can do something to their cute dog........

Another interesting point is that BIOS can detect memory address 0x472 to know this is a warm or cold reboot we want. If the reboot mode is 0x1234 that is a warm reboot. If reboot mode is 0 that is cold reboot...;-)

BTW, this is my desktop wallpaper recently...;-)
And I want to say that openSUSE rocks!!
Also the president Ma of TW rocks!!

This hot girl is named 王思平 and she is a famous model of TW!!
I personally love her bott....XD

Jaunty was released on 4-23-2009

2009-04-28T14:57:00.007+08:00

Ubuntu 9.04 Jaunty was released on 4/23/2009.
Although I don't like this distro, I still have to install and evaluate it on my company's NB products...Orz

This release of Ubuntu implement the ext4 file system which was improved from previous ext3 fs type.

First, I have tasted the Ubuntu 9.04 Netbook Remix edition. This is the first formal release for Ubuntu distro which got a optimized system on netbook products. With a different kind of "desktop" interface which can be more useful with touch screen device such as MID.

Basically, Ubuntu have made the OS image for LPIA and/or netbook product, but they didn't formal announce it such as 9.04. I downloaded those non-announced image and installed it to give it a try. I have to say the truth that, those releases suck..damn..!! Very low performance and laggy response of the whole system, it's very terrible........

Now, finally 9.04 netbook remix with ext4 supported, it is just OK to use. Even I still see error messages of built-in Intel i915 / drm driver.

Linux kernel 2.6.29 is released

2009-03-25T22:12:00.004+08:00

Cool~!

On 2009-03-23 the Linux kernel 2.6.29 is out there~!
Let's find out how many cool stuffs' inside this release.

Message from Linus: click here.

For the changes in 2.6.29: click here.

BTW, I love this new logo - Tuz, it's cute.

Code::Blocks IDE in openSUSE 11.1-build and install

2009-03-13T12:48:00.006+08:00

Again, today I will try to build and install the Code::Blocks, an open source and free IDE for programming, in my own openSUSE 11.1 box...;-)

1. First, I need wxWidgets supported in my openSUSE 11.1 box. In a terminal console:
$ zypper search wxWidgets

I got some information about it as:
正在載入儲存庫資料...
正在讀取已安裝的套件...
S | 名稱 | 摘要 | 類型--+-----------------------+-------------------------------------------------------+-----------
| wxWidgets | C++ framework for cross-platform development | 套件
| wxWidgets | C++ framework for cross-platform development | srcpackage
| wxWidgets-compat | WxWidgets Compatibility Package | 套件
| wxWidgets-debuginfo | Debug information for package wxWidgets | 套件
| wxWidgets-debugsource | Debug sources for package wxWidgets | 套件
| wxWidgets-devel | Include files and libraries for wxWidgets development | 套件
| wxWidgets-docs | WxWidgets html documentations | 套件
| wxWidgets-gl | OpenGL add-on for wxWidgets | 套件
| wxWidgets-gl-compat | OpenGL add-on for wxWidgets | 套件
| wxWidgets-lang | Languages for package wxWidgets | 套件

So, I decided to install some packages:
$ sudo zypper install wxWidgets wxWidgets-compat wxWidgets-devel wxWidgets-lang

System will automatically check your box and tells you about how many new packages would be installed.
Just type your passwoed and hit 'y' to go.

2. Get Code::Blocks source code via SVN. (Or you may want to check this page first):
$ svn checkout svn://svn.berlios.de/codeblocks/trunk svn-trunk/

This will try to checkout the latest Code::Blocks source code trunk to the directory called "./svn-trunk/"
Revision of my code that checked out is 5483.

3. Just read the file inside your truck directory called "BUILD", and now I can build and install it:
$ cd svn-trunk/
$ ./bootstrap

Output message is:
libtoolize: putting auxiliary files in `.'.
libtoolize: copying file `./ltmain.sh'
libtoolize: Consider adding `AC_CONFIG_MACRO_DIR([m4])' to configure.in and
libtoolize: rerunning libtoolize, to keep the correct libtool macros in-tree.
libtoolize: Consider adding `-I m4' to ACLOCAL_AMFLAGS in Makefile.am.

$ ./configure --enable-contrib

Output message is a little bit weird but seems fine to go on......
I checked the help page of ./configure but there is no "--enable-contrib" option......:
…
...
...
configure: WARNING: unrecognized options: --enable-contrib
*************************************************
* Code::Blocks source tree has been configured. *
*************************************************
You can now build Code::Blocks by issuing 'make'.
When the build is complete, become root and install
it by issuing 'make install'.

Finally, we can:
$ make
$ sudo make install

All good to install and play with it...;-)

Install RTL8187SE PCI-e wifi card on Ubuntu 8.10

2009-02-04T16:50:00.005+08:00

I just got a new platform with the RTL8187SE wifi card (PCI-e).
Damn it doesn't work on Ubuntu 8.10 with native kernel module rtl8180.ko........Orz
I think maybe the device ID is not supported (my device ID is 10ec:8199).

I try to grep some useful infomation about source code to install by myself.
And, now it works!
See below..;-):

1. Download the source code deb package from here. (name: rtl8187se-source_1023.0928.2008_all.deb)

2. I DO NOT want to use dpkg to install this, so I use command dpkg --extract to extract the src only.

!! CHECK THE README FILE INSIDE THE PACKAGE BEFORE YOU DO ANY THING !!

3. $ cd ./modules/rtl8187se

4. $ sudo ./makecln; sudo ./makedrv

5. $ sudo ./wlan0up (this is a test for new kernel module r8180.ko that we just made.)

If there is no problem,

6. $ sudo ./wlan0down; sudo ./install

System will automatically depmod -a and reboot. Done..;-)
BTW, the kernel for this process is 2.6.27-7-generic.

Cable with RJ45 jacks how-to

2009-01-16T16:39:00.003+08:00

RJ45 jack (tenon faces to ground..):

Normal:
Both jacks: 白橘橘白綠藍白藍綠白棕棕

Cross-over:
a jack: 白綠綠白橘藍白藍橘白棕棕
b jack: 白橘橘白綠藍白藍綠白棕棕

How to install p7zip 4.61 - UNIX/Linux version of 7-zip on openSUSE 11.1

2009-01-08T16:51:00.008+08:00

I install it because I love it...;=)
Again, I'd like to compile it via source code by myself.

BTW, if you are NOT an advanced user I think, you should try the pre-built binary package instead of compiling it by yourself.

Well, if you now stay with M$ Windows I still recommend it. Check it out.

1. Get source code package from here. Currently, latest verson is 4.61.

2. My box is openSUSE 11.1 x86_64, T9500 CPU...;-). So I just un-tar and cd into the directory:
$ cd ./p7zip_4.61/
$ sudo zypper search -d yasm

I got output such as follows:
正在載入套件庫資料...
讀取已安裝的套件...

S | 名稱 | 摘要 | 類型

--+----------+--------------------------------------------------+-----
---| yasm | YASM is a complete rewrite of the NASM assembler | 套件

3. Install yasm:
$ sudo zypper install yasm

I got output such as follows:
正在載入套件庫資料...
讀取已安裝的套件...
正在解決套件相依性...

將會安裝下列新的套件：
yasm

全部下載大小： 240.0 K。完成操作後，將增加 654.0 K 的使用。
要繼續嗎？ [是/否]: y
正在取出套件 yasm-0.6.2-1.75.x86_64 (1/1), 240.0 K (已解開 654.0 K)
正在取出： yasm-0.6.2-1.75.x86_64.rpm [完成]
正在安裝： yasm-0.6.2-1.75 [完成]

4. Modify ./makefile.machine. Here is my makefile.machine:
$ cp makefile.machine makefile.machine.backup
$ cp makefile.linux_x86_asm_gcc_4.X makefile.machine
$ vim ./makefile.machine

I modify ./makefile.machine as following RED words for my T9500 x86_64 system:
#
# makefile for Linux (x86, PPC, alpha ...)
#

OPTFLAGS=-O

ALLFLAGS=${OPTFLAGS} -m64 -s \
-D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE \
-DNDEBUG -D_REENTRANT -DENV_UNIX \
$(LOCAL_FLAGS)

CXX=g++ $(ALLFLAGS) -DHAVE_GCCVISIBILITYPATCH -fvisibility=hidden -fvisibility-inlines-hidden
CC=gcc $(ALLFLAGS)
CC_SHARED=-fPIC
LINK_SHARED=-fPIC -shared
ASM=yasm -f elf -m amd64

LOCAL_LIBS=-lpthread
LOCAL_LIBS_DLL=$(LOCAL_LIBS) -ldl

CPU=x64
OBJ_CRC32=$(OBJ_CRC32_x86_64)

5. Of course, make it!
$ make clean; make all2; sudo ./install.sh
Note that if you need 7zr just replace above "make all2" with "make all3".

Done...;-)

How to install mplayer and browser plug-in on openSUSE 11.1 w/ Firefox 3.0.5

2009-01-04T23:51:00.017+08:00

OK first, HAPPY NEW YEAR...X-D

Today I just want to install a movie player for my openSUSE 11.1 box.
Well, mplayer is always my favorite one...;-)

BTW, I know there are several repositories which provided pre-built package.
Hmm..., I think I just like to do this by myself instead of using package which I don't trust (security reason...:-P).

So, here is a very simple memo for installation steps.
Basically I want to do 2 tasks:
A.) Install mplayer with GUI enabled (gmplayer).
B.) Install mplayer plug-in for Firefox web browser.

=======================
==== A. Install mplayer ====
=======================
1. Get source code. Use following command to do so:
$ svn checkout svn://svn.mplayerhq.hu/mplayer/trunk mplayer
There is a directory called "mplayer" which will be created under current directory for storing source code of mplayer. So you should have the write permission for this dir. Note that if you use this way to get source code (svn) you will have libav* from FFmpeg inside already.

2. Get proper binary codecs package. Go to here and look for your CPU arch. type and download a binary codecs package. Note the if you used a wrong codecs package, the compile and install process will be failed. In my box I use Linux x86 20071007.

3. Get mplayer skin for enabling GUI feature. Go to here and scroll down the page to skins section. Select any skin you prefer and download it. I use Blue.

4. Install the binary codecs package. In my box I install into /usr/local/lib/codecs which is the default path that mplayer will look for. Issue followinig commands:
$ tar jxvf essential-20071007.tar.bz2
$ cd ./essential-20071007/
$ sudo mkdir -p /usr/local/lib/codecs
$ sudo cp -rfv ./* /usr/local/lib/codecs

5. Compile and install mplayer w/ GUI enabled. Issue following commands:
$ cd ./mplayer/
$ ./configure --enable-gui --enable-menu
$ make
$ sudo make install

6. Create a symbolic link for ttf. It will be used for mplayer OSD. In my box I issue following commands:
$ cd /usr/local/share/mplayer
$ sudo ln -s /usr/share/fonts/truetypes/uming.ttf subfont.ttf

7. Install the skin for GUI (gmplayer). Issue following commands:
$ tar jxvf ./Blue-1.7.tar.bz2
$ cd ./Blue/
$ sudo mkdir -p /usr/local/share/mplayer/skins/default
$ sudo cp -rfv ./* /usr/local/share/mplayer/skins/default/

8. It's done!! Let's play!!...;-) Just type mplayer or gmplayer with a movie file.

=======================
==== B. Install plug-in ====
=======================
1. Since the browser plug-in path between every distros is a little bit different. So we need to identify where we should copy our mplayer plug-in to. An easy way to do this just search a plug-in which was already installed for your browser. In my box I type:
$ sudo find / -name npwrapper.libflashplayer.so -print

Then I know the plug-ins are in the path /usr/lib64/browser-plugins/.

2. Grep the source package from here.

3. Un-tar the package and read the README/INSTALL first. Then I know that I have to install Mozilla devel files or Gecko SDK to compile mplayer plug-in. Issue following commands to install Gecko SDK package on openSUSE 11.1:
$ zypper search -d gecko
$ sudo zypper install mozilla-xulrunner190-devel

4. Change to mplayer-plugin's directory. And compile it:
$ cd ./mplayerplug-in-3.55/
$ ./configure --with-gecko-sdk=/path/to/gecko-sdk (replace /path/to with the correct value on your system)
$ make

5. Install *.so and *.xpt files, close your Firefox now and:
$ sudo cp mplayerplug-in*.so /usr/lib64/browser-plugins/
$ sudo cp mplayerplug-in*.xpt /usr/lib64/firefox/components/

6. Bingo! Launch your Firefox and enter "about:plugins" in the URL box then hit Enter, check the mplayer plug-ins are all installed...;-)

Additional packages for compiling kernel on Ubuntu 8.04.1

2008-12-23T17:24:00.002+08:00

I apt-get installed follows:

build-essential libtool makedepend libncurses5-dev linux-kernel-devel libcunit1-ncurses-dev

Of course this is for make menuconfig.

I also found some pages but I never tried their way:
http://ubuntuforums.org/showthread.php?t=56835
https://help.ubuntu.com/community/Kernel/Compile
http://www.howtoforge.com/kernel_compilation_ubuntu

This is not my first time to compile Linux kernel.
But on Ubuntu...hmm, yes it is...囧rz

Actually I like Redhat/Fedora, Mandrake more........= =|||

Compiling SiS X driver on Ubuntu 8.10

2008-12-19T16:46:00.003+08:00

In order to compile the X driver for SiS chip, I have to apt-get install following packages additionally:

build-essential
libtool
makedepend
libgl1-mesa-dev
x11proto-*
xserver-xorg-dev
libdrm-dev

Then, of course, ./configure; make.
Cd to ./src/.libs/, you can see your sis_drv.so...;-D

Now I just try to port this version to support another chip...Orz

(You want the source code for SiS X driver...hmm...I think it is not a difficult task to you, right?)

RealPlayer 11 on Linux

2008-12-19T10:02:00.002+08:00

I try to install RealPlayer 11 on my DELL Vostro 1310 with Mandriva 2008.1, kernel 2.6.24.7-desktop. All things are fine. why this player? Hmm......somtimes I feel the .RM files are really pains in my ass and I don't have time to search for other codec or player for it. So..you know that...;-)

Download RealPlayer here. You can choose .BIN format (personally recommand) or even RPM/DEB package if you want.

Just cd to where you put the file and issue commands with root permission:
(optional step) $ chmod a+x RealPlayer11GOLD.bin
$ ./RealPlayer11GOLD.bin

Follow on-screen instruction to answer some questions. Use the default installation path is ok on my system (/opt/Real/RealPlayer).

Note that the RealPlayer is not able to start with zh_* locales so that I have worte a simple script to change my $LANGUAGE dynamically to en_US.UTF-8 (or en_US*) to launch it...;-)

Install NV driver under Ubuntu 8.10

2008-12-19T09:52:00.002+08:00

I have also tested Ubuntu 8.10 with kernel 2.6.27-7-generic.
Basically all steps are almost the same as 8.04.1 but only this one:

For step 2, only you have to do within Ubuntu 8.10 is "S30gdm".

I use nVidia 180.11 Beta Linux X driver.
Currently official release is 177.82.

DLink DI-604 and DIR-605

2008-12-11T14:02:00.002+08:00

Two DLink broadband network IP switchies in the same place, 1st is DI-604 and second one is DIR-605:

1. Use DI-604 to connect to internet via xDSL/Cable, and connect DIR-605 "WAN port" to DI-604 LAN port. Disable DIR-605 NAT function. Just treat it as a regular "wireless" hub/switch. This is not a good idea but only workable solution...Orz...And, in this case you can NOT use MAC filter to lock up some MAC addresses you want, because the NAT of DIR-605 is disabled........

2. Use DI-604 to connect to internet via xDSL/Cable, and connect DIR-605 "LAN port" to DI-604 LAN port. Disable DIR-605 DHCP server function only and change DIR-605's IP to differ from DI-604. This is more easier and you may use MAC filter function on DIR-605!...;-)

Check it out here.

Install NV driver under Ubuntu 8.04.1

2008-12-08T18:54:00.002+08:00

The process is very similar to install it under Mandriva 2008.1.

1. Prepare driver; install "build-essential" package.
2. Use root permission to rename 2 file under /etc/rc2.d:
-a.) S20nvidia-kernel (I rename it to __S20nvidia-kernel)
-b.) S30gdm (I rename it to __S30gdm)
3. Reboot and use root permission to change to your driver's dir and run:
root# sh NVIDIA-xxxxxxx.run
4. No not use nvidia-xconfig, just manually edit /etc/X11/xorg.conf. (refer to Mandriva's process).
5. Use root permission to rename back the files /etc/rc2.d/__S20nvidia-kernel and /etc/rc2.d/__S30gdm to their original file name.
6. Reboot it and enjoy it...;-)

Install NV driver under Mandriva 2008.1

2008-12-08T17:01:00.002+08:00

You need to download NV driver for Linux. Go to here.
Note that you should always refer to file "latest.txt" for latest released driver.
This is for 32bit system.

Setup your Mandriva 2008.1 box with a workable network.
Then you have to install additional packages for building NV driver.
I suggest that you use the Mandriva Software Manager to intall those additional packages because it will automatically select all related packages for you.

1. gcc;
2. binutils;
3. automake;
4. make;
5. makedepend;
6. gcc-c++;
7. kernel devel package according to your kernel version (uname -r);

After all done, use root permission to edit /etc/inittab.
Replace run level index from 5 to 3. Reboot.

Login and su to root, change dir to the driver package's directory.
Then issue command:
root# sh NVIDIA-xxxxxxxxx.run

Follow instructions on screen.
If you are asked to download pre-compiled interface from NV FTP site just say "NO".

If you are asked to run "nvidia-xconfig" to modify xorg.conf just say "NO".
We will do it later (for safety reason..).

After driver building finished, you should issue a reboot.
Login again and use root permission to edit /etc/inittab and switch run level index back to 5. Save it.
Use root permission to edit /etc/X11/xorg.conf, within section "Device" ="Configured video device" and add a line (or edit it):
Driver "nvidia"

Save it and reboot.
Now you should see the NV driver load up and will show you a logo screen of the NV driver...;-)

Load up a customized DSDT on Ubuntu 8.04.1

2008-11-06T13:49:00.003+08:00

I got some pains in my original DSDT so that I decide to load up a customized version of DSDT.aml.
Here is what I have to do in my Ubuntu box.

1. Backup vmlinuz and initrd images under /boot (use root permission to keep owner/group):
$ sudo cp /boot/vmlinuz-$(uname -r) ~/vmlinuz-$(uname -r)
$ sudo cp /boot/initrd.img-$(uname -r) ~/initrd.img-$(uname -r)

2. Copy your DSDT.aml to proper dir and reconfigure your initrd image (case-sensitive):
$ sudo cp DSDT.aml /etc/initramfs-tools/DSDT.aml
$ sudo dpkg-reconfigure linux-image-$(uname -r)

3. Edit GRUB boot menu (depands on your demand, optional):
$ sudo vim /boot/grub/menu.lst
- Find out the booting kernel you want, and append a parameter "acpi=force" after the line starting with kernel.

4. Save it and reboot. Use "dmesg |more" to check if there is any error happened for new DSDT.

Sierra Wireless USB modem (GSM/UMTS) setup in Ubuntu 8.04.1

2008-10-21T11:12:00.007+08:00

I am now tried to verify two 3G modules to my recent project. That is the Sierra mini card module which was attached on USB interface. Such as:

In Ubuntu w/ kernel 2.6.24-19-generic I can compile and install new driver v1.3.1b without any problem. I have tried 2 different modules MC8775 and MC8790, both of them work very good.

There are some points you may interested in it:
1. After driver installed, you can use command "modinfo sierra" to check version and it should be v1.3.1b.
2. By default, Sierra driver uses pppd to connect. And use /dev/ttyUSB* to communicate with device. If you received an error such as modem initialization failed or connect script failed, you can try to modify the string "/dev/ttyUSB0" (this is default value) in your /etc/ppp/peers/gsm script (this is default Sierra provided script path of pppd). Just try to use another ttyUSB[1..n] which you can use command "ls /dev/ttyUSB*" to know how many ttyUSB you can try.

3. If you can dial out successfully and get IP configuration but you can not surf on internet with browsers, just try to cp /etc/ppp/resolv.conf to /etc/resolv.conf.

BTW, if you use the 3G device from Acer Peripherals Inc. (now BENQ Corp.), you don't need additonal kernel driver module but only usbserial.ko. First use lsusb to check the VID:PID. In my case that is 04a5:0468, so I issue command to load up kernel driver module for this device: sudo modprobe -v usbserial product=0x0468 vendor=0x04a5. Second, you just need to wait for few "seconds" for driver to take effect and create proper device nodes for you (/dev/ttyUSB[0~2]). Then you can use pppd to dial out just like the way you do on Sierra devices. Note that this 3G module's communication port is /dev/ttyUSB1 for pppd.

How to uninstall Trend Micro OfficeScan client w/o IT password

2008-10-20T17:16:00.005+08:00

Sometimes I want to re-install Trend Micro OfficeScan client from my office PC.
But due to my IT dept.'s rule it requires a password to uninstall the client.

How can we do to "drive-around" it? ;-)

Basically, we can use a little bit trick to do so.

Use "services.msc" command to open the service list. And then find out all "OfficeScanNT*" services and STOP them. You can see the Trend Micro OfficeScan icon in the system tray will disappear.

Open your registry editor and find out HKEY_LOCAL_MACHINE\System\Current Control Set\Services, and look for following subkeys:
Ntrtscan
Tmfilter
Tmlisten
TmPreFilter
TmProxy
tmtdi
VSApiNt
TM_CFW (if you have this one)

Delete above subkeys (CAUTION! BE CAREFUL WITH THIS OPERATION!).

Find out the registry key HKEY_LOCAL_MACHINE\Software\TrendMicro, and look for following subkeys:
OfcWatchDog
NSC
Pc-cillinNTCorp
OfficeScanCorp (if you have this one)
RemoteAgent (if you have this one)

Delete above subkeys (CAUTION! BE CAREFUL WITH THIS OPERATION!).

Find out HKEY_LOCAL_MACHINE\Software\MicroSoft\WINDOWS\Current Version\Run, and look up the value: OfficScanNT Monitor,

Delete above value (CAUTION! BE CAREFUL WITH THIS OPERATION!).

Delete the entry in your "Start" -> "Programs" -> "Trend Micro OfficeScan xxx", and then issue a system reboot. After reboot delete C:\Program Files\Trend Micro\OfficeScan Client or C:\OfficeScanNT, Then also you can delete the file C:\tmuninst.ini if you have it.

Ok, now you are ready to re-install your virus scaning application ;-)

Mandriva 2009.0 just released

2008-10-11T13:21:00.004+08:00

In the early today, I just received a system notice about Mandriva 2009.0 just released (..OK, I forgot to shut it down...Orz). The major updates were: kernel 2.6.27, X.org 4.3, KDE 4.1.2 (and 3.5.10 in /contrib), Gnome 2.24, FireFox 3, and OpenOffice.org 3.

Basically, I plan to stay in my 2008.1 w/ 2.6.24 kernel for a while. Because I know the new release should be in a "POST-BETA" stage but not a very stable one. My 2008.1 just works fine with my DELL Vostro 1310 and I don't want to screw it up because I really don't have time to fix any issues that caused by the new 2009.0 release...;-)

I plan to update my system maybe 1 or 2 months later.

For details and the errata (this is very important in my opinion..) of Mandriva 2009.0, you can refer to here.

Mandriva One Spring 2008.1 bugs

2008-10-04T17:27:00.002+08:00

My DELL Vostro 1310 was installed the Mandriva One Spring 2008.1.
I don't need to install extra drivers for my unit because Mandriva Linux is a very good distro for hw supporting. And also it is very easy to use for everyone including experts and newbies.

Unfortunately, I still got some bugs with my box.

1. System memory was found only up to 884 MB (it should be 2 GB in my box).
2. Intel wifi card 4965AGN got response timeout issues that sometimes can not access to the AP.

For 1, this is a kernel issue so that I switch my default kernel build from kernel-desktop586 to kernel-desktop. You can use $ uname -r to make sure your kernel build currently on your system. And this issue was fixed by this kernel build switching. Based on the official errata, you need to also install some kernel driver modules in order to let the switching more stabler if you have those drivers for kernel-desktop586 arch.

For 2, just add a line into /etc/modprobe.conf:

options iwl4965 disable_hw_scan=1

Official errata you can refer to here.

Confusing hard disk device in Debian 4.0 R4a - continued

2008-10-01T17:39:00.002+08:00

My Debian 4.0 R4a box has another problem that, after installation of the system the resume device was changed!! Normally it was the swap file system, and it will be defined during the installation process and wrote into initramfs (/boot/initrd-img-xxx).

I think the installer of Etchnhalf is really buggy!
He recognized my hdd and odd as:
/dev/hde -> hdd
/dev/hde1 -> 1st partition, boot
/dev/hde2 -> logical
/dev/hde5 -> swap
/dev/hdg -> odd

And he wrote it into /boot/grub/menu.lst for kernel boot parameter as root=/dev/hde1. Of course the resume device in initrd-img will be /dev/hde5, the swap.

OK, now I complete installation process and reboot my Debian 4.0 R4a box. Guess what!?

I can not boot into system because the stupid installer makes a WRONG device recognation!...:-(
And now the boot device and root file system can not be found and mounted........

The actual device map of my box should be:
/dev/hdc -> hdd
/dev/hdc1 -> 1st partition, boot
/dev/hdc2 -> logical
/dev/hdc5 -> swap
/dev/hde -> odd

Now, I need to modify my /boot/grub/menu.lst and /etc/fstab to correct this stupid problem which was caused by Debian installer.
And I need to issue command $ dpkg-reconfigure uswsusp to re-generate the /boot/initrd-img-xxx........:-(

Confusing hard disk device in Debian 4.0 R4a

2008-10-01T14:40:00.003+08:00

This is weird..= =|||

I have installed Etchnhalf many times but the HDD device file changed every time?!
Maybe this time it is /dev/hdc in istallation process, but it will be changed to /dev/sda after installation done.
Then I hit Ctrl+Alt+Del to reboot, it may be changed back to /dev/hdc or more worse /dev/hda..
I think this is related to kernel driver of ide-generic and SATA.
Maybe they are fighting to each other for controling my IDE channel...:-(

And one more weird thing is that the kernel parameter "quiet".
Basically, there will be a parameter "quiet" added in the end of kernel boot command line in modern Linux distro.
I don't like this parameter because I totally can not monitor my Linux box if it displayed some error message when booting up.

In this case I shut "the quiet" down so that I can:
1. To see error messages of the HDD device file changed.
2. To fix the ide port error when GRUB wants to load kernel up (Fucking weird! WHY??????).
Why this error is related to kernel boot parameter "quiet"??????

Debian 4.0 R4a + JMB38x can not find out my MSP cards

2008-09-23T15:42:00.007+08:00

In my Debian system (Debain 4.0 R4a w/ kernel 2.6.24-etchnhalf.1-686), I have a JMicron multimedia card reader (JMB38x which supports MMC/xD/MS). After I installed JMircon's driver I found my Debian can not auto detect the MSP cards but xD cards are OK.

The first point I checked is my udev process. I think the hotplug function is abnormal for MSP cards. Therefore I modify the following 2 file under /etc/udev: hotplug.rules and permissions.rules.

For /etc/udev/hotplug.rules just modify it as following pic w/ root permission:

For /etc/udev/permissions.rules also modify it as following pic w/ root permission:

Save them and reboot machine, now I can see my MSP cards in my system as /dev/mspblk0 and /dev/mspblk0p1 device files because the mspro_block module was successfully load-up by udev.

Of course, I need to mount it manully to access MSP cards such as $ sudo mount -t auto -o rw /dev/mspblk0p1 /media/tmpdisk.
That is OK for me 'cause I don't need auto-mount. ;-)

Looking for FLV player? Check this out

2008-09-15T20:08:00.003+08:00

Basically, there are so many flv players in the world.
I would like to recommand the following one: Wimpy Desktop FLV Player.

Why this one? Because of:
1. It is free.
2. The flv video screen size can be adjusted.

http://www.wimpyplayer.com/products/wimpy_standalone_flv_player.html

You can directly drag and drog a file to player to play. Or you can click the folder icon to open a file in your computer (flv, swf...etc). Click on the earth icon can let you connect to a file to play which was located in somewhere on this planet!

Usplash sometimes caused a black screen

2008-09-12T20:42:00.002+08:00

My Ubuntu is 8.04.1 box with kernel 2.6.24-19-lpia.
After Usplash screen sometimes I got a black screen and I can not do anything but poweroff..
This happened in MP units with Intel Poulsbo platform but not found in PVT units..it's strange!
The fail rate in production line is around 50%.

I delete the kernel parameter "splash" and everything goes fine.
End user will see the very ugly verbose text and maybe they will feel nervious with it.
But anyway, they can use this box....

I think this bug also related to Poulsbo gfx driver in Linux and X.
I have tried a patch on launchpad but doesn't worked for me.
https://bugs.launchpad.net/ubuntu/+source/usplash/+bug/81474

RHAS4+PXE+DRBL+CloneZilla setup

2008-09-07T15:19:00.005+08:00

This is a memo of my experience to setup a DRBL+CloneZilla server.
I use offline intallation method.

(A). Base OS and DRBL installation
Linux server installation (RHAS4 for example):
1. Use default installation. DISABLE SELINUX and firewall (iptable).
2. If necessary, use REISERFS, otherwise use EXT2/EXT3.
3. Don’t install office, multimedia, game, video/sound…etc.
4. Manually install some missing RPMs: perl-Digest-SHA1, dhcp, ypserv, tftp-server for RHAS4 (with root permission).

$ rpm –ivh perl-Digest-SHA1-xxx.rpm
$ rpm –ivh dhcp-xxx.rpm
$ rpm –ivh tftp-server-xxx.rpm
$ rpm –ivh ypserv-xxx-rpm

5. Setup network interface eth0:1 if necessary (IP alias if your only have 1 NIC).
Modify /etc/sysconfig/network-scripts/ifcfg-eth0:1.
Reference site: http://drbl.nchc.org.tw/one4all/desktop/ip_alias.txt
6. Setup HDD DMA=1 if necessary. (/etc/sysconfig/harddisks).

Download and make new kernel RPM:
1. Make new kernel RPM to built-in support (not module) ide-generic in our case and some NIC drivers.
2. Reference site: http://www.howtoforge.com/kernel_compilation_fedora

Download and install DRBL related RPMs (use root permission):
1. Download DRBL GPG key: http://drbl.nchc.org.tw/GPG-KEY-DRBL
2. $ rpm --import GPG-KET-DRBL
3. Download DRBL main RPM:
http://drbl.nchc.org.tw/one4all/desktop/download/stable/RPMS/drbl-current.i386.rpm
4. $ rpm –ivh drbl-current.i386.rpm
5. Download extra DRBL RPMs:
http://drbl.sourceforge.net/drbl-core/i386/RPMS.drbl-stable/
6. $ rpm –ivh xxx.rpm to install all extra RPMs (CloneZilla RPM should be the last one to install).
Before you perform next section, it is better to do a reboot.

(B). DRBL setup – OFFLINE mode (use root permission)
1. Type command: $ /opt/drbl/sbin/drblsrv-offline –k (kernel RPM) to use specified kernel RPM or $ /opt/drbl/sbin/drblsrv-offline –s `uname -r` to use same kernel as server used.
2. DRBL will check all necessary RPMs automatically. If need, press Ctrl+C to cancel DRBL setup process. Follow on-screen instruction to answer question.
3. Type command: $ /opt/drbl/sbin/drblpush –i to push settings to clients node. Follow on-screen instruction to answer question.
Reference site: http://drbl.nchc.org.tw/one4all/desktop/examples/
4. Type command: $ /opt/drbl/sbin/mknic-nbi –t initramfs –k (Kernel version that
client used) to make initramfs boot image for clients.
5. Copy image "folder" to /home/partimag directory (default path). For example: /home/partimag/20080812-beta3-image

(C). Setup image restore option and PXE menu (use root permission):
Unicast restore:
1. Type command: $ /opt/drbl/sbin/drbl-ocs -b --nogui -s -x -r [-p command] [--run-postrun-dir] startdisk restore 20080812-beta3-image "hda"
-b: batch mode. No question aksed.
--nogui: without GUI during restore process.
-s: Don’t detect client HW.
-x: Try to enable full-duplex in UDPcast. (only work with switch, not hub.)
-r: Try to resize partition after restore process finished.
-p [reboot | poweroff | command]: Post-action after restore process finished.
--run-postrun-dir: Excute program or shell script under /opt/drbl/share/ocs/postrun after restore precess finished.
--run-prerun-dir: Excute program or shell script under /opt/drbl/share/ocs/prerun after restore precess finished.
[startdisk | startparts]: CloneZilla restoring mode.
[restore | multicast_restore | save]: CloneZilla restoring type.
20080812-beta3-image: Image "folder" under /home/partimag.
“hda”: device or partition in image to restore.

Multicast restore:
1. Type command: $ /opt/drbl/sbin/drbl-ocs -b --nogui –s -r [-p command]
--clients-to-wait 1 --time-to-wait 10 --max-time-to-wait 60 [--run-postrun-dir]
startdisk multicast_restore 20080812-beta3-image "hda"
-b: batch mode. No question aksed.
--nogui: without GUI during restore process.
-s: Don’t detect client HW.
-r: Try to resize partition after restore process finished.
-p [reboot | poweroff | command]: Post action after restore process finished.
--clients-to-wait NUM: Minimum numbers of client to wait before restore process starting.
--time-to-wait NUM: Minimum seconds to wait before restore process starting.
--max-time-to-wait NUM: Maximum seconds to wait before restore process starting (to avoid CloneZilla service time-out).
--run-postrun-dir: Excute program or shell script under /opt/drbl/share/ocs/postrun after restore precess finished.
--run-prerun-dir: Excute program or shell script under /opt/drbl/share/ocs/prerun after restore precess finished.
[startdisk | startparts]: CloneZilla restoring mode.
[restore | multicast_restore | save]: CloneZilla restoring type.
20080812-beta3-image: Image "folder" under /home/partimag.
“hda”: device or partition in image to restore.

Misc:
*** Open every shell script to read and study so that you can know every thing that DRBL+CloneZilla server does! ***

PXE boot menu is: /tftpboot/nbi_img/pxelinux.cfg/default
Postrun dir is: /opt/drbl/share/ocs/postrun
DRBL conf dir: /etc/drbl and /opt/drbl/conf
DRBL official site: http://drbl.nchc.org.tw/ or http://drbl.sourceforge.net/
FAQ: http://drbl.nchc.org.tw/faq/ or http://drbl.sourceforge.net/faq/

Ghost 11.5 little problem

2008-09-07T12:29:00.003+08:00

Oops!
I found that using Ghost 11.5 -id to clone Linux system, partiton table may be modified in some special situations.

I got my Linux partiton table as follows:
/dev/sda1, boot flag, ID 83, Linux native
/dev/sda3, ID 82, Linux Swap

But after Ghost cloning it became to:
/dev/sda1, boot flag, ID 83, Linux native
/dev/sda2, ID 82, Linux Swap

It caused Linux can not enter S4 (hibernation) status because Linux doesn't know where is the swap file system!

I use a very stupid way to fix it: re-write MBR!
Since this problem was happened in factory production line and, any action to boot into OS to fix partition table is not allowed :-(

Simply prepare a small tool called "MBRUtilD.exe" to save / restore MBR section.
I use it to create a workable MBR w/ partiton table, and then use tftp (via PXE) to transfer to clients for restoring.

"MBRUtilD.exe /R=(MBR file name)" to restore MBR.
And option "/S=(MBR file name)" to save MBR.
Just run it right after Ghost cloning.

It rocks of course! ;-)

Use Ghost 11.5 to create / restore Linux file system image

2008-09-06T15:37:00.001+08:00

Ghost is a very powerful tool to create / restore OS images.
But Ghost doesn't support GRUB boot loader in MBR (my experience with old Ghost) if your HDD got Linux file systems.
Usually you need to manually fix and re-install boot loader after you cloning a machine.

I just try Ghost 11.5 with option -id and this does support GRUB in MBR without any problems!
Really cool!

Option -id use sector-by-sector method to create image. So that you won't miss any data on your HDD including MBR, boot sector and so on.

I also recommend to use CloneZilla to create / restore Linux OS image. This is a great utility with open source license. Just check: http://www.clonezilla.org/