Thanks to everyone that come to visit or even leave a comment. It’s really good to know someone is still visiting my small space.

The reason I stopped writing blog is because I’ve get involved in a really really big project, and got quite busy. The project is still not finished, but is in a good shape. I think I’ll get sometime to update this blog.

The reason I came to visit my own blog is because I was looking for an old article . And the reason I looked for that old article is because I’ve got myself a new PC. With the summer getting hotter and hotter, my old AMD3000+ machine halts several times a day. I know switching to a bigger fan might help, but, well, it is getting slow and the halting is an good excuse for a new one

So I go get myself an I5-750 machine, with 4G RAM. It’s much faster than the old one. It is so fast that a buildworld in a FreeBSD on FreeBSD virtualbox client is faster than the real 3000+ machine. I also got myself a Dell 2410 and setup a two-monitor desktop. Finally I can have two 1024×1200 window sitting side by side, it looks really nice

So much for the news. I’ll get back and update my blog because again, I found it’s a good for future work if I wrote the tips down.

Enjoy the life and live with passion.

Share This

First of all we’ll need to setup FreeBSD as a network bridge.

This should return “bridge0″ if this is the first bridge you created.

And we’ll add the two interface used on the bridge. It’s “rl0″ and “vr0″ in my case:

Now bring the bridge up. Note the two interface need to be up too.

If you want an IP address of the box, assign it to the bridge

Now the bridge should work. To make dummynet work, we’ll need to recompile the kernel:
add

to the kernel config file and rebuild it. Then add

to /etc/rc.conf.

Make sure you have filewall_type=”open” in /etc/rc.conf and/or options IPFIREWALL_DEFAULT_TO_ACCEPT in your kernel config file before you do the reboot if you are doing this remotely. Or you’ll block yourself out.

Now enable ifpw on the bridge

OK, we can now add a pipe:

Note the “1000″ in the command line. Do a

first and choose a good place for the pipe.

Now it’s time to shape the traffic. To add some delay:

or add some package lost:

Reference:

A nice tutorial

The FreeBSD hand book about bridging

The FreeBSD hand book about IPFW

dummynet(4)

ipfw(8)

Share This

I find this out because the cracker created a new user on his server, and runs the password scanner on his server to gain access to more servers.

The cracker deleted the C source code of the scanner, but left some shell scrip running. And here is how it works: The scanner got a pretty large username/password match database, and the scanner will run through this database against all the computers online. Pretty simple, hmmm? And with some simple tricks, like scan those machines has SSH connection with the current machine, it get pretty efficient.

The thing scares me most is that they’ve got a really large database. Not only the usual words in dictionary or simply phases like “letmein”, I saw things like “mima” which is the Chinese pronunciation of “password”. There are also some password that I think from other langnuages.

And my friend confirmed his sys admin used to use some really simple password. So, use SSH2 key auth whenever you got a chance, and when you have to use a password, nomatter on SSH or not, never think your password is complicated enough.

Share This

The most problem of setting up Vsftpd with file based virtual user on FreeBSD is that we don’t have pam_userdb. So we’ll use /usr/ports/security/pam_pwdfile instead.

So after installing the vsftpd and pam_pwdfile ports, do the following:

(1) in /etc/pam.d, make a file named “vsftpd” and put the following lines in it:

(2) make a password file with the apache htpasswd utility:

(3) make a local user with “nologin” as the shell and disable password login. I use “vsftp” for this.

(4) in /usr/local/etc/vsftpd.conf, disable anonymous access and enable local user access, and then add the following:

Note the value of “guest_usrname” is the local user name you made in step (3)

(5) make a directory /usr/local/etc/vsftpd, and write user specific configurations in a file named after the user’s login name. For example, someone with full write access will have something like this:

(6) It should be OK now, add vsftpd_enable=”YES” into /etc/rc.conf as usual, and restart vsftpd to see if everything works fine.

Share This

I really like this small one cause I can carry it everywhere and save me some time while I’m walking or waiting.

It support file/folder visiting of the files, and not support MP3 tag and organized with artiste, album, etc like Apple IPod. This is the reason I prefer it over the Sandisk Clip. Because most of the time I’ll use it to listen podcasts or audio books.

However there is a problem: It won’t sort the files. I just can’t find any options to sort the files in the menu system or the user user manual. Then I thought it might sort the files by their access time, or create time, so I wrote mp3 reorg , which will change the atime/ctime of all the files so the fils should appear earlier will get an earlier atime/ctime. However it turns out this little mp3 don’t obey that too, they just read the FAT table and don’t do any sort! Finially I have to wrote another one, mp3 copy, to copy the files in the order I want.

Just put them here in case anyone could find them useful. Note “create time” is a dos/windows only thing, we don’t have this on (most) Unix systems. If your mp3 use atime (access time) or ctime (changed time), you could use the “mp3reorg” script on Unix platform too, just comment out the lines with “win32″ in it. The “mp3 copy” script works cross platform, though.

Share This

To move the mail server to another server, and lost mails as less as possible, here is my steps:

- Set your DNS server so your domains get a lower TTL. Do this as early as possible if you have a high TTL before.

- Setup mail server on the new server, use the same settings on the old server, and make sure everything works.

- Sync all the mails from the old server to new server. This takes a long time

- Sync again, this should take a shorter time

- Stop the old mail server

- Do a final Sync

- Start the new mail server

- Update your DNS so the mx record point to the new server

- Done. Check the status and make sure things are OK.

Some note:

- This is for situations that we only have one mail server. If we already have a backup mail server, things will be different.

- We don’t use the priority of mx record here, cause we don’t want in-consistency between the two servers

- Most mail servers will retry for at least 24 hours if they can’t deliver a mail (either because we close down the old mail server or their DNS is not updated yet and can’t find the new server) so as long as TTL is small enough, we are good and no mail should be lost.

Share This

However after several days of usage I found I can only use 2G~3G memory on typical days. The rest are free or intact. This might because I’m still using the config files I tweaked for my old notebook, where I have only 256M memory. I was thinking once to switch back to FreeBSD-i386 to get the Binaries like Google Earth or win32-codecs, or something that couldn’t compile on amd64 like audio/beep or textproc/sdcv.

But I’ve got another idea: I could use the extra RAM to build a RAM disk, and mount it on /tmp. Thus I could take advantage of this much faster /tmp. Of cause all the content on a RAM disk is lost if I do a reboot, but I already have

in my /etc/rc.conf so I got nothing to lose.

And it turns to be easier to setup than I thought. Just add one line

into /etc/fstab and do a reboot. Among other things the “-s” option tells how big I want the RAM disk to be. And I think 512M is good for me. Also a RAM disk won’t take all the 512M memory all the time. This is just the max size of the RAM disk. And when I have only a little files in /tmp, it takes only a little RAM. Check md(4) for all the details.

Share This

Searching the web and mailists and I found some reports that for people don’t use Xmodmap, the CapsLock key and the NumLock key won’t update the LED on the keyboard on Xorg7.3. Checking my box and I found I also get this problem. I guess this is something related, comes with the Xorg7.3.

Fortunately, besides Xmodmap, we have another option to set the CapsLock to Ctrl, or switch CapsLock and Ctrl. For me, I added

in the “InputDevice” section of my xorg.conf file, and now I can use the CapsLock as Ctrl key as before.

For people who want to switch these two, use “ctrl:swapcaps” instead. If you want to try it out, you can also use the setxkbmap command, for example:

Check the man page for more options.

Share This

Contents

• Why
• What do we want
• Accessible with a single key stroke
• Available on all virtual desktops
• No Title, No border
• Transparent background
• Not shown on task-bar, task switch, or Alt-Tab
• And finally, nice animation

Why

Well we already have many quake like terminals already:

• Kuake
• Yakuake
• Tilda
• Stjerm

But what if you don’t like loading the whole KDE libraries to get
this? Or, what if all those programs lacks some features you really
like, like a transparent background to your root window with a nice
tint color?

Today we’ll turn a general purpose terminal, like the Terminal
program for the Xfce project, or rxvt-unicode, to a quake like
terminal, with Fvwm.

I think I don’t need to give too much introduction to Fvwm: If you
are reading this, there is a good chance that you are already using
it

What do we want

First of all we’ll need to find out what do we want from a Quake like
terminal. Here’s the list:

• Accessible with a single key stroke
• Available on all virtual desktops
• No title, No border
• Transparent background
• Not shown on task-bar, task switch, or Alt-Tab
• And finally, nice animation

In this article we’ll use the Terminal program as an example. And
we’ll set its title to ” quick_terminal”.

Ready? Let’s rock!

Accessible with a single key stroke

To make the quick terminal Accessible via a single key stroke, we’ll
need two things:

1. If the terminal is not running, start one
2. If the terminal is already running, make it at the right place and focus on it

Here’s the code:

We’ll explain this line by line.

The first two line is how we define functions: AddToFunc says to add
the following command to a function, with the function name following
it. If the function does not exists, then create one. The
DestroyFunc is a safe keeper: It removes any, if there is any,
functions with the same name. There’s no overload kind of thing in
Fvwm configure file. You may wonder why not just a “DefineFunc”. The
answer is that Fvwm configuration is really flexible: You can add
instructions to one function in different places.

Then we come to the line:

There’s a lot to talk about if you are new to Fvwm configuration
files:

The + I at the beginning says the command in this line will be
invoked Immediately. This is what we need most of the time.

The None (”"quick_terminal”, CurrentDesk) follows is a condition
statement in Fvwm. It means “Run the following command when there’s
no window has the title “quick_terminal” in the current desk”. None
is the verb here, “when there is no… do…”. And the stuff between
the brackets are the conditions. We are going to make our terminal across
all the virtual desktops of Fvwm, so we add CurrentDest here, to be
more efficient: Fvwm won’t need to search all the desktops for this
title name now.

Then Exec exec Terminal -T quick_terminal. Exec is a general
command in the Unix world, and it means what you are thinking . We
use double Exec to make sure the following command returns
immediately. This is recommended in the Fvwm manual. And finally, the
command we want to run: Terminal -T quick_terminal. This starts a
Terminal instance with the title set to quick_terminal. You can
replace this with whatever terminal program you like, for example,
urxvt -T quake_terminal.

The following several lines are easy to understand:

First we release our quick_terminal from Iconify status. So the window
is not minimized anymore. Then we Resize it and Move it to a location
we like. I’m assuming you are using a 1024×768 screen and this will
bring the terminal in the middle of your screen, with 800×600
pixels. The Raise command makes the terminal window appears on top
of all the other windows, and the Focus command will put the
keyboard focus to this window, so what we types next, will be
transferred to our quick_terminal. The final Break command is,
well, a break, like in any other programming languages. We’ll talk
more about this later.

The following hide_quick_terminal command is rather simple: We
first set the Iconify status of our quick_terminal on, so it
disappears and lost the keyboard focus. Then we use Next Raise to
set the keyboard focus to the window we are working on before calling
our quick terminal. Next here is another conditional command in
Fvwm. We don’t use any conditions (stuffs between brackets) here so
it just selects the next window in the stack.

The last function, show_or_hide_quick_terminal, calls the two
functions above in different conditions: ThisWindow is another
conditional command. It tests if the current windows matches the
conditions following. So the first two line of this function says: If
I’m working on a window named “quick_terminal”, call
hide_quick_terminal, else call raise_or_fire_quick_terminal. The
final NoWindow conditional command matches when we are not on any
window. This is useful when, for example, you just started Fvwm and
haven’t got any program running.

OK. We’ve got all these set. The final step would be, make our primary
function, show_or_hide_quick_terminal, be invoked with a simple
key stroke. We’ll use the F1 key, as an example:

Here Key starts a key binding. F1 is the key we want. The
following A and N needs some explanation: A defines the context
of this key binding. That is, where this key binding is available? on
an application window? a root window? or the side bar (border) of a
window? We use A here so it’s available in all conditions. Then the
modifier N. If you want Ctrl-F1, Alt-F1, or Shift-F1 instead
of just F1, you can put C, A, or S here, accordingly. This
time we think the quick terminal is so useful that we just want an
F1, so we put a N here, which means “no modifier”".

And here’s why we add a Break statement at the end of the first two
functions: show_or_hide_quick_terminal is the function get
called. If we don’t add a break some where, all the three statements
in it will be called one by one. Say if we are on the quick_terminal
window. The first condition matches, and hide_quick_terminal get
called. And now we are at another window, and the second condition
matches, so raise_or_fire_quick_terminal get called too, and our
window re-appears. That’s definitely not what we want.

Available on all virtual desktops

This is easy with Fvwm:

Style is the Fvwm command controlling how a window
behaves. Following by the window name (title) and the
behavior. Sticky is the style this time. It means the window will
be available on all virtual desktops.

No Title, No border

Here is the code:

The first line makes the quick_terminal window has no title bar, no
handles, and no borders, as you can guess from the name. Handles is
the tiny triangle on the lower-right corner that you can drag to
resize the window. As we want a fixed width and height of this quick
window, and to make sure there’s no borders anyway, we turn this off
for our terminal.

The second line is a safe keeper, again. In case you have some
settings that turns on Handles and/or Borders for all the windows,
this will make the width zero.

Transparent background

This is really not a window manager’s work. Well, at least until
Compiz and Beryl comes out.

All I have to say is to choose a terminal emulator program that have
good transparent background support. There are a lot: Terminal,
rxvt-unicode, aterm, eterm, etc. I may talk about how to set your
background window to be compatible with these programs in a later
article.

Not shown on task-bar, task switch, or Alt-Tab

Fvwm uses the FvwmTaskBar module for the MS windows like “task
bar”. If you don’t like it and don’t use it, fine, nothing will appear
on the task bar, cause there is no task bar

If you are using task bar, here’s how to hide our terminal from it:

The first line add style WindowListSkip to our terminal. This
prevent the window from appearing in the menu that the WindowList
command created. This is the menu that most people bind to Alt-Tab.

And the second line, is a Fvwm option line for FvwmTaskBar. This
tells FvwmTaskBar not to show the window with a WindowListSkip style,
on the task bar.

With these two lines, our always available terminal won’t pollute our
window switching window and task bar.

And finally, nice animation

Hmmm, want to look cool? Let’s add some animation here We’ll make a
pull-down animation cause this is what all the other quake like
terminals have. And you’ll find out soon that Fvwm can do much more.

The Fvwm module that handles animation is FvwmAnimatie. To start it,
add the following line to your Fvwm configuration file:

With this started, there is a good chance that you can find an
“animation” item on your “settings” menu. If not, go to your Fvwm
Console and issue the command:

then the menu popups up, where you can choose the animation effects, line width,
delay times, etc. Because we haven’t turn animation on for our little
command when we make our terminal Accessible with a single key
stroke, you can’t see the effect right now. Just make sure the
effect is Frame here.

Now we need to add two:

statements to those two functions. This means “send an animate
command to the FvwmAnimate module”. And it takes 8 parameters: x1,
y1, w1, h1, x2, y2, w2, h2. The first four defines a rectangle where
the animation starts, and the last four defines a rectangle where the
animation ends.

For us, we need to add:

before:

in function raise_or_fire_quick_terminal and:

before:

in function hide_quick_terminal.

Now we are all set. Restart your Fvwm and press F1. Do you see the
magic? Use your imagination here, for example, you can make a
rectangle rolls out from the upper left corner when you want the
terminal. And this, is left for you as an exercise

Share This

Well I need to handle some PHP upgrade on a Debian server some days ago, and it turns out to be a nightmire…

Debian is famous for its “apt-get” package management system, and I thought it would be easy. However, it turns out all the packages are binary and have exact version numbers in dependency. To make it worse, the previous sysadmin uses both sarge and etch packages. Now if you want to install something from sarge, it complains some dependence in etch is too high and won’t continue, and if you want to install something from etch, well, some dependences got a version too low.

And I tried the “smart” “apt-get dist-upgrade” command, and it removed my bind9! All sites going down because of this. And to make it worse, I can’t install it again because of the sarge/etch mix. And finially I have to deinstall many packages and re-install them, and stick to sarge.

I know this is not a pure problem on Debian, the previous sysadmin should be blamed mostly, but, hi, the binary package system with hard coded version numbers makes all these possible… And I have to say, compile from the source rocks! The FreeBSD ports system rocks!

Share This