Ilap's tech

Downgrade iPhone 3G with no working Home Button from iOS 4.2.1 to 3.1.3

2011-07-03T00:39:00.000-07:00

Disclaimer:
First, I take no responsibility for any trouble caused by this info and make no guarantees of any kind.
So, please be very carefull when you want use any info from this post.
SO, YOU'RE USING ANY INFO IN THIS POST AT YOUR OWN RISK!

Let's rock.:)

Yesterday afternoon I was given an iPhone 3G, unfortunately that phone come /w faulty home button. I thought it will good for my wife or one of my children but /w no home button it's useless as it stucks in the application when the app's activated and there is no way to come out to main screen

As I did not want spend a cent for a fix I thought a software solution like Activator or similar should do the trick. But, unfortunately Activator just runs on a jailbroken iPhone. I did not want to get that iPhone jailbroken but had not have any other option as I could not find any similar app in App Store. So, I am a really newbie in this kind of things so started to read the Internet how can it be done.

After a lot of reads I got more confused as this is not as easy as it looks.
In general there are two type of jailbreaks are exist:

Tethered (non-permanent, so you need to jailbreak again after power cycle)
Untethered (permanent)

Also there are several jailbreak methods out there but all the methods are depends on several factors such model, model number, iOS version, modem FW and so on.

So, lets start /w the phone specification:
Model: iPhone 3G
Model nr: MB496X
iOS version: 4.2.1 (8C148)
Modem FW: 05.15.04 (Unlocked)

An iPhone /w iOS 4.2.1 the jailbreak method in general would be the following:
1. Choose the proper jailbreak method
2. Put the iPhone to the DFU (Device Firmware Upgrade) mode.
3. Jailbrake it.

In brief, the DFU or Device Firmware Upgrade mode allows all devices to be restored from any state.
The DFU is a big problem as you might be know that to put an iPhone into the DFU mode you need to use the Sleep/Power and the Home button together.
There is no way to put my iPhone to DFU with the method above.
So, tried to find some other methods and just tow of them came up as a possible solution:
1. LAST DFU Software
2. True Hardware DFU Mode Automatically

Unfortunately the LAST DFU does not work /w my 3G iOS 4.2.1 and it's always entering to Recovery Mode instead of DFU mode. I think it's a known issue /w this latest iOS.

Anyway, on some older iOSes (3.1.2, 3.1.3, 3.2 and some others), some untethered jailbreaks (like Spirit) can be very smooth as the iPhone not needs to be entered into DFU mode to get it JB-d. See details for the vulnerability of iPhone, iPhone 3G and iPhone touch (NOT for iPhone 3GS and iPod Touch 2G): http://theiphonewiki.com/wiki/index.php?title=S5L8900

So, if I can get the DFU mode somehow why do not downgrade the phone to some old original iOS and try some or more untethered and non-DFU methods instead of the slow and risky HW DFU Mode and jailbrake on the current 4.2.1 iOS every time when I fail? In that case I just need enter to DFU once, restore it to an official iOS and try the user-space method anytime I want.

So, I thought that my steps would be the following:

Download the official iPhone 3G 3.1.3 and iPhone 3G 4.2.1 IPSW files
Put iPhone into DFU mode
Restore it /w an official iOS 3.1.3 using iTunes
Kicks the iPhone out of Restore Mode loop after it's upgraded using RecBoot
Do some user-space (non-DFU mode) jailbreak /w Spirit
Install Activator /w Cydia and reassign the Home Button to Sleep/Power or Volume button.

Please be advised that this procedure will destroy all of your stored data and installed Apps as the 4.2.1 iOS backup is not compatible /w downgraded 3.1.3 iOS.

THEREFORE YOU WONT BE ABLE TO RESTORE YOUR STORED DATA FROM A BACKUP.

I did not matter for me as I did not have any stored data there, but maybe somebody has a bunch of app installed and songs purchased.
So, please consider the above before proceed as I do not know how should you get the data back.

I have the Mac OS X 10.6.8 /w iTunes 10.2.2 installed on my MacBook Pro, which is hardly booted as I use Ubuntu instead, but we need to do the following on a Mac OS X as I do not have any Windows PC to test it and the iTunes is not installed in my Ubuntu.

You need to download the following apps/files:

Download the original iPhone 3G 3.1.3 and iPhone 3G 4.2.1 IPSW files

I downloaded these two original IPSWs (IPhone SoftWare) from this link.
The 3.1.3 is the target iOS version on my iPhone and the 4.2.1 needs for trying to put my iPhone onto DFU mode as it needs to be on the same or above version as the current one as Apple does not allow to downgrade any iDevice (iPod, Iphone and iPad) from the installed version.
So, after I downloaded them I did the following steps.
The following command must be typed in a terminal.
In Mac OS X, the terminal application can be found under Utility, so click Application (in Finder) and then Utility and Terminal.
A cmd like window will appear and you can start typing the following:
$ cd ~
$ mkdir IPSW
$ mv Download/iPhone*ipsw IPSW
$ cd IPSW
$ ls -rtl
...
-rw-r--r-- 1 ilap ilap 239139281 2011-07-02 15:32 iPhone1,2_3.1.3_7E18_Restore.ipsw
-rw-r--r-- 1 ilap ilap 338579563 2011-07-02 16:00 iPhone1,2_4.2.1_8C148_Restore.ipsw

Put iPhone into DFU mode

I read several articles and some of the tries were successful and some were not. I am certain that if somebody is doing the following steps on the same type of iPhone 3G /w the same specification I have then it will success.

This trick is based on an official Restore for iPhone 3G using an actual or newer version official IPSW file which is corrupted/modified to do the trick.

The IPSW files are just a zipped files, so we just need:

unzip the IPSW,
modify some file of the unpacked ones,
repack the IPSW /w the modified file and
do an iTunes Restore /w this corrupted file.

And hopefully during the Restore process (at least at the end of that) the iPhone will stuck in the DFU mode and we can do an Official FW Restore process /w the official 3.1.3 IPSW file. As I told the iTunes cannot do a downgrade in Restore Mode as it check that /whihc version of the iOS is currently installed on the iPhone, but in DFU mode there is no any iOS is installed (as it's completely failed) therefore we can install any officially supported IPSW files in it.
At the end of downgrade process the iTunes come up /w an 1015 error and will left the iPhone on the Recovery Mode loop, but it can be easily kick out of this mode using Recboot v2.2 MAC OS X or similar.

So, lets get down to busines:
1. Corrupt the IPSW
$ cd ~/IPSWs # To the downloaded files
$ ls - iP*
-rw-r--r-- 1 ilap ilap 338579762 2011-07-01 22:40 iPhone1,2_4.2.1_8C148_Restore.ipsw
-rw-r--r-- 1 ilap ilap 239139281 2011-07-02 15:32 iPhone1,2_3.1.3_7E18_Restore.ipsw
$ cksum *
1805666746 239139281 iPhone1,2_3.1.3_7E18_Restore.ipsw
1672942545 338579762 iPhone1,2_4.2.1_8C148_Restore.ipsw

Save the original for later check
$ mkdir orig
$ cp -pr iPhone1,2_4.2.1_8C148_Restore.ipsw orig

Check the files
$ unzip -tq iPhone1,2_4.2.1_8C148_Restore.ipsw

No errors detected in compressed data of iPhone1,2_4.2.1_8C148_Restore.ipsw.

$ unzip -lqq iPhone1,2_4.2.1_8C148_Restore.ipsw */LLB*img3

68980 2010-10-12 17:52   Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3

$ unzip -l iPhone1,2_4.2.1_8C148_Restore.ipsw */LLB*img3

Archive: iPhone1,2_4.2.1_8C148_Restore.ipsw
Length      Date    Time    Name
--------- ---------- -----   ----
68980 2010-10-12 17:52   Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
---------                     -------
    68980                     1 file

Extract the LLB file
# unzip -x iPhone1,2_4.2.1_8C148_Restore.ipsw */LLB*img3

Archive: iPhone1,2_4.2.1_8C148_Restore.ipsw
inflating: Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3

Copy the LLB file for later check
$ ls -rtl Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
-rw-r--r-- 1 ilap ilap 68980 2010-10-12 17:52 Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
$ cksum Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
1912212578 68980 Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3$ cp Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3 orig.IPSW

Corrupt some part of that file (some random bytes on offset 0x240)
This below is a one line command and not two lines.
Also, the conv=notrunc is mandatory to prevent the file shrinking after write.
# dd if=/dev/urandom of=Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3 bs=16 count=10 seek=36 conv=notrunc10+0 records in
10+0 records out
$ ls -lrt Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3 orig.IPSW
-rw-r--r-- 1 ilap ilap 68980 2011-07-02 15:55 orig.IPSW
-rw-r--r-- 1 ilap ilap 68980 2011-07-02 15:55 Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
$ cksum Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3 orig.IPSW
1753812497 68980 Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
1912212578 68980 orig.IPSW
As you can see above the sizes are same but cksum, which is fine

Repack the corrupted file
$ zip -fv iPhone1,2_4.2.1_8C148_Restore.ipsw Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
freshening: Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
    zip warning: Local Entry CRC does not match CD: Firmware/all_flash/all_flash.n82ap.production/LLB.n82ap.RELEASE.img3
     (in=68980) (out=42523) (deflated 38%)
total bytes=338820395, compressed=338574217 -> 0% savings

Compare the orig and the corrupte IPSW file
$ ls -lrt iPhone1,2_4.2.1_8C148_Restore.ipsw orig/iPhone1,2_4.2.1_8C148_Restore.ipsw
-rw-r--r-- 1 ilap ilap 338579762 2011-07-01 22:40 orig/iPhone1,2_4.2.1_8C148_Restore.ipsw
-rw-r--r-- 1 ilap ilap 338579563 2011-07-02 16:00 iPhone1,2_4.2.1_8C148_Restore.ipsw
$ cksum iPhone1,2_4.2.1_8C148_Restore.ipsw ../Orig/iPhone1,2_4.2.1_8C148_Restore.ipsw
3489769154 338579563 iPhone1,2_4.2.1_8C148_Restore.ipsw
1672942545 338579762 orig/iPhone1,2_4.2.1_8C148_Restore.ipsw

Restore it /w an official iOS 3.1.3 using iTunes

So, start the restore proceduer as usual:

Fully charge the iPhone
Power off the iPhone (slide to power off)
Connect the iPhone to the PC using USB cable
Power on the iPhone
if iTunes does not launch when the phone is ready start the iTunes
Select your iDevice in the left Tree of iTunes
Wait for syncing and backup, when it's done
Press ALT and click to Restore
An Open Filed Dialog will be appeared and choose the corrupted IPSW (IPSW/iPhone1,2_4.2.1_8C148_Restore.ipsw)
Click OK and the restore is started.

It will start the Restore process as usual and put the phone to Restore Mode (an Apple logo in the middle and a progress bar below that), so It will crash at some stage (about 5 minutes later) close to the end of Restore, and the iTunes will come up /w error (error 37), and the phone will stuck /w black screen (this means it stuck in the DFU mode, Hoooooraaaay).

So, we can start the REAL restore /w the official iOS 3.1.3 file.

Close iTunes
Unplug the phone from computer
Start iTunes
Plug the phone back to computer
Press ALT and click to Restore and choose the file
An Open Filed Dialog will be appeared and choose the official IPSW (IPSW/iPhone1,2_3.1.3_7E18_Restore.ipsw)
Click OK.

It will start the Restore process as usual and put the phone to Restore Mode (an Apple logo in the middle and a progress bar below that).
At the end of the Restore the iTunes come up /w the Error 1015 and the phone will entered the Recovery Mode loop (an USB cable and an iTunes icon).

Kicks the iPhone out of Restore Mode loop after it's upgraded using RecBoot

That's error message above is normal, so we just need the phone kicked out form that Recovery Mode loop which can be done with RecBoot. It's simple just download it and start it and it would say that a "Recovery Device Connected" and just click to Exit Recovery and wait until your iPhone is booted.
So, that's all. If everything went smooth than you have an iPhone with official 3.1.3 iOS installed.

Do some user-space (non-DFU mode) jailbreak /w Spirit

So, it's as easy as the RecBoot was previously, just download it from the link in this post and click to "Jailbreak" button and wait until it's doing his job.
Install Activator /w Cydia and reassign the Home Button to Sleep/Power or Volume button.

When your phone is jailbreked, you can find the Cydia icon. Click to that icon and will do some package updates and some other stuffs.
When it's done you can search for Activator in the Cydia repository and install.
After it's installed you could find an Activator menu in the bottom of global Settings (main scrren Settings button). So, what I did:

Click to Settings on the iPhone's main screen
Find and Click to the Activator at the bottom of the Settings page
Click to the Anywhere
Scrolled down to the "Volume Buttons" and
Clicked to "Down Short Hold"
Scrolled down to "System Actions" and select Home Button

Ind this case the Home Button will be assigned to the Short Hold on the Down Volume Button. From this point if you push down the volume button for short time (1-2sec) and release it, it will trigger a Home Button push, which is what I just wanted.

NOTE: During the Cydia update my iPhone's frozen just one circle was on the middle of screen which was just circling. After ten minutes I tried to power off the phone, but I could not as the Power button was inresponsive. I tried to do a soft reset (hold the power button for a while) but no success. So, I just started the RecBoot and clicked to the Enter Recovery mode button, as soon as I clicked the iPhone, it's rebooted and entered into the Recovery Mode (iTunes icon and an USB cable sign). I waited a bit replugged the iPhone, startent the Recboot and clicked to the "Exit Recovery". After this the iPhone is rebotted as normal.

Conclusion

The phone is working properly /w this settings and have not observed any problems issues /w it.

Galaxy Live Wallpaper like Compiz plugin for Ubuntu 10.10 (Maverick Meerkat)

2010-10-09T17:42:00.000-07:00

I think I do not need to introduce the live wallpapers of Android and you guys should know everything about it.
It was a love at first sight when I read about it in the Nexus One live wallpapers post.
Especially the Galaxy Live Wallpaper was my favorite one.
I thought that would be very nice if somebody could implement the Live Wallpapers or at least the Galaxy Live Wallpaper in the upcoming Ubuntu Maverick (10.10) somehow.
But, there was a big silence in the Internet, so I had no option but to try to create some similar program if I want to see it in my Ubuntu (10.10) desktop.

As the Compiz is available in almost all of the Linux distros and has a plugin framework, therefore I thought that would be great if I could implement some similar as a Compiz plugin. To tell the truth, when I started this project I had no any experience in OpenGL and Compiz plugin programming.
Fortunately, I finally managed to create that plugin for my Ubuntu 10.10 Desktop. I created a screenshot, see below, to show you how it looks like.

My aim was to release this plugin by 10 of October, 2010 (10.10.10) the release date of the Ubuntu 10.10 (Maverick Meerkat), and it seems that I have achieved it.
I have also created and uploaded a short desktop video to Youtube (see below the embeded video) to show you all what to expect from it. Unfortunately, it is very low quality one, but should enough.

I have uploaded the source code to Launchpad for the brave and a PPA package (ppa:ilap/lwp) to easy install for the lazy ones.
One of the most missing part of the source code is the CREDITS file, where I am going to put on all of the people name who inspired me to get this done.
And please consider that is my first Compiz plugin/OpenGL app therefore it could be full of bugs and hidden sprites.:)

To tell the truth, I do not know whether it will work on your system or not, as I have not had any chance to test it in some other notebooks/PCs, but it works very well on my notebook.:) .
When the Galaxy is enabled, the compiz just uses about 8% of CPU resources on my 2.80GHz Core2 Duo notebook.

So, here comes some brief information about my system and if you have some similar then it should work on your one too.

$ grep model\ n /proc/cpuinfo
model name    : Intel(R) Core(TM)2 Duo CPU     T9600 @ 2.80GHz
model name    : Intel(R) Core(TM)2 Duo CPU     T9600 @ 2.80GHz

$ glxinfo | grep -i render
direct rendering: Yes
OpenGL renderer string: Mesa DRI Mobile Intel® GM45 Express Chipset GEM 20100330 DEVELOPMENT x86/MMX/SSE2

$ glxinfo | grep version
server glx version string: 1.4
client glx version string: 1.4
GLX version: 1.4
OpenGL version string: 2.1 Mesa 7.9-devel
OpenGL shading language version string: 1.20

So, this VGA chipset, which is relative new, supports OpenGL 2.0. I would say if your chipset supports OpenGL 2.0 (and some extensions) then it should work properly. If it does not support then the plugin can be very slow as some GL hardware function will be passed to the software renderer.
So, in that case you should expect that compiz will use at least 90% of CPU and provide just 2-3FPS.:) I think this can be fixed by some advice come from some high-experienced OpenGL/Compiz plugin programmers as I am not one of them.:)

Install and use

The installation should be relative easy task, just follow the steps below.
BUT, keep in mind this plugin is created for Ubuntu 10.10 (Maverick Meerkat) and should NOT work in the previous releases:

Step one: Install the plugin

1. Add my PPA to the repository.
   $ sudo add-apt-repository ppa:ilap/lwp
   $ sudo apt-get update

2. Install the package
I named it compiz-plugins-livewallpapers as my aim is to make my source/PPA as a container for the LiveWallpapers like plugins.

   $ sudo apt-get install compiz-plugins-livewallpapers compizconfig-settings-manager

If somebody decided to install it from the source he/she should follow the following steps:
   $ bzr branch lp:compiz-plugins-livewallpapers
   $ cd compiz-plugins-livewallpapers
   $ sudo apt-get install `grep "Build-Depends" debian/control | \
sed -e 's@^.*: @@' -e 's@([^(]*)@@g' -e 's@,@@g'`
   $ ./autogen.sh
   $ ./configure --prefix=/usr
   $ make && sudo make install

Step two: Configure the plugin

You need to change your Desktop background to my installed background one.
Right click on the Desktop -> Change Desktop Background -> click to "Add" button ->
type "/usr/share/backgrounds" to "Location" field -> Press ENTER -> select Space_by_ilap.jpg
-> click to "Open" button -> click "Close" button

After the background is changed you need to enable (if it's not already enabled by the install process) the "Galaxy Live Wallpaper" plugin:
Gnome Panel -> Systems -> Preferences -> CompizConfig Settings Manager
The ccsm window will appear and you can see on the left a new category called "Live Wallpapers".

Click on the "Live Wallpapers" category on the left.
Enable Galaxy Live Wallpaper

Step three: Toggle it

Just press "Super" (Windows logo button) and F4 buttons to toggle it on or off.
If you are lucky then enjoy, if you are not then ask for help...

So, good luck and have fun...

I hate Monster (Apple) Inc.

2010-07-05T06:13:00.000-07:00

I hate Apple Inc. inspite of I love some of its products (Mac OS X, Macbooks, Mac Mini etc. etc.) as those and some other stuffs/gadgets are of good quality and so on, BUT, I have just read that the Apple was granted a patent on the Cover Flow interface.
I did not know that this patent was handed at all and am a bit confused as the 1st reference to the Cover-Flow-Like UI is back to 29th of December in 2004, but on that time that page and its contents was covered by Creative Common 2.0 NC-SA. Anyway, I am not a license guy at all just wondering how come that the Apple could granted that patent. Also as I remember that I saw a bunch of Cover-Flow-Like works in the past, which was related to Java 2D/3D.
Anyway, here comes the info I could find in the Net. Of course, I can be wrong but this mess should be cleaned by someone.

1. 29th of December in 2004: Initial thoughts (Common Creative 2.0 BY-NC-SA)
All of my respect for the author...
Reference: Dissatisfaction Sows Innovation. Keep in mind it was a mockup and not a real app.
Picture:
2. 26th of June in 2005: Joplin made its debut at "NetBeans Day", at JavaOne 2005
Reference: Joplin MP3 Player and it has an LGPL license.

3. 1st of July in 2005: Joplin blog post
Picture:
4. July 7th, 2005: 3D Effect with Java2D?
Keep in mind in this post was the 1st reference that this demo (CDShell) was run in a Mobile phone (Motorola e680) in that time (2005). So, iPhone CoverFlow????
Picture:

5. July 09, 2005:A Music Shelf in Java2D.
Picture:

5. August 13th, 2005: An independent Mac OSX developer created the CoverFlow App
This is the initial version of that app which was purchased by Apple Inc. on ~11th of September in 2006
Picture:

6: 09th of November in 2005: Sneak Preview of Next Demo
XVID Video and Picture

7. 11th of September in 2006: Cover Flow and all of the Andrew Coulter Enright's IP related to the idea were purchased by Apple Inc. and CoverFlow was integrated into iTunes 7.0
Picture:
8: 28th of June in 2007: Patent filed
9: 28th of April in 2010: Patent granted

And finally, back to the past.

I was very keen on the Sun's Project looking glass (See pictures below) in late 2003 and early 2004, whihc was released under GPL v2.0 on 30th of June in 2004.
Ooooh, Does the Dock seem to be similar with the Apple Leopard's one?
When do you think the Leopard was released with its 3D Looks Dock? Yep, on 26th of October in 2007.
Oh, you think it was easier to the Java guys to renew the Apple's old style dock from Apple Mac OS X Public Beta (2000). Yes, you could be right, but do not forget the Sun's and others real dock in CDE Env back to 1993.

References:
Decemer 12th, 2003: Sun's Project Looking Glass
March 25th, 2004: (news) Project Looking Glass - making Linux 3D!

That's all Folks!

Add proxy support to openconnect

2009-12-17T01:20:00.000-08:00

I just realized this afternoon that the Cisco VPN access to our Private Net has been decommissioned for two days. Ooooh... it's really a big problem because I'm working from home tomorrow and the Security team hasn't responded yet to my question relating to the new Cisco AnyConnect VPN (SSLVPN) client access.
I've asked them why they didn't upload the Linux package to the ASA flash and declare it in the config as an AnyConnect image in order to be able to connect to our Corporation Private Net from a Linux client.
Yep, I use LinuxCOE (COE: Common operating environment) on my Office laptop (http://linuxcoe.sourceforge.net/).

So, I had to try to use use the openconnect instaed using the original Cisco's AnuConnect SSL VPN Client to test the new SSLVPN connection to our VPN.
After I downloaded it I realized that the openconnect doesn't support any connection through an https proxy. Unfortunately, I have to use a https proxy to access to the webvpn, but the openconnect doesn't support it. One more bugger.:)

So, I had no option but to write a patch for the openconnect to support proxy.
Long story sort... here comes the patch.

$ cat openconnect-2.12_proxy.diff


diff -Naur openconnect-2.12/http.c openconnect-2.12.ilap/http.c
--- openconnect-2.12/http.c 2009-12-08 05:40:34.000000000 +1300
+++ openconnect-2.12.ilap/http.c 2009-12-18 00:28:54.000000000 +1300
@@ -637,3 +637,63 @@
 
  return uagent;
 }
+
+void openconnect_parse_proxyurl(struct openconnect_info *vpninfo, char *url)
+{
+ char *proxy_port;
+ char *proxy = url;
+ vpninfo->proxy = proxy;
+
+ /* 
+  * if the format is hostname:port then we cut off the port from name
+  */ 
+ proxy_port = strchr(proxy, ':');
+
+ if (proxy_port) {
+  *proxy_port = 0x0;
+  proxy_port++;
+  
+  vpninfo->proxy_port = proxy_port;
+ }
+ /* else the default one (1080) which is already set  will be used  */
+}
+
+int process_http_proxy(struct openconnect_info *vpninfo, int ssl_sock)
+{
+ char buf[MAX_BUF_LEN];
+ int buflen;
+
+ if(!vpninfo->proxy) {
+  return 1;
+ } else if (ssl_sock < 0) {
+  return -EINVAL;
+ }
+
+ bzero(buf, MAX_BUF_LEN);
+ sprintf(buf, "CONNECT %s:%s HTTP/1.1\r\n", vpninfo->hostname, vpninfo->port);
+ sprintf(buf + strlen(buf), "Host: %s\r\n", vpninfo->hostname);
+ sprintf(buf + strlen(buf), "User-Agent: %s\r\n", vpninfo->useragent);
+ sprintf(buf + strlen(buf), "Proxy-Connection: keep-alive\r\n");
+ sprintf(buf + strlen(buf), "Connection: keep-alive\r\n");
+ sprintf(buf + strlen(buf), "Accept-Encoding: identity\r\n");
+ sprintf(buf + strlen(buf), "\r\n");
+
+ vpninfo->progress(vpninfo, PRG_INFO, "Connecting to the proxy %s on port %s\n",
+   vpninfo->proxy, vpninfo->proxy_port);
+ vpninfo->progress(vpninfo, PRG_TRACE, "==>%s<==\n", buf);
+
+ buflen = write(ssl_sock, buf, strlen(buf));
+ if (buflen < 0) {
+  return -EINVAL;
+ }
+
+ bzero(buf, MAX_BUF_LEN);
+ buflen = read(ssl_sock, buf, MAX_BUF_LEN);
+
+ if ((buflen < 0) || strncmp(buf, "HTTP/1.1 200", 12)) {
+  vpninfo->progress(vpninfo, PRG_ERR,"Proxy connection failed!");
+  return -EINVAL;
+ }
+
+ return 1;
+}
diff -Naur openconnect-2.12/main.c openconnect-2.12.ilap/main.c
--- openconnect-2.12/main.c 2009-12-08 05:40:34.000000000 +1300
+++ openconnect-2.12.ilap/main.c 2009-12-17 17:25:55.000000000 +1300
@@ -68,6 +68,7 @@
  {"syslog", 0, 0, 'l'},
  {"key-type", 1, 0, 'K'},
  {"key-password", 1, 0, 'p'},
+ {"proxy", 1, 0, 'P'},
  {"user", 1, 0, 'u'},
  {"verbose", 0, 0, 'v'},
  {"version", 0, 0, 'V'},
@@ -109,10 +110,11 @@
  printf("  -i, --interface=IFNAME          Use IFNAME for tunnel interface\n");
  printf("  -l, --syslog                    Use syslog for progress messages\n");
  printf("  -U, --setuid=USER               Drop privileges after connecting\n");
- printf("      --csd-user=USER           Drop privileges during CSD execution\n");
+ printf("      --csd-user=USER             Drop privileges during CSD execution\n");
  printf("  -m, --mtu=MTU                   Request MTU from server\n");
  printf("  -p, --key-password=PASS         Set key passphrase or TPM SRK PIN\n");
  printf("      --key-password-from-fsid    Key passphrase is fsid of file system\n");
+ printf("  -P, --proxy=PROXY[:PORT]        Set proxy server\n");
  printf("  -q, --quiet                     Less output\n");
  printf("  -Q, --queue-len=LEN             Set packet queue limit to LEN pkts\n");
  printf("  -s, --script=SCRIPT             Use vpnc-compatible config script\n");
@@ -182,6 +184,9 @@
  memset(vpninfo, 0, sizeof(*vpninfo));
 
  /* Set up some defaults */
+ vpninfo->proxy = NULL;
+ vpninfo->proxy_port = OC_DEF_PROXY_PORT;
+ vpninfo->port = OC_DEF_SSL_PORT;
  vpninfo->tun_fd = vpninfo->ssl_fd = vpninfo->dtls_fd = vpninfo->new_dtls_fd = -1;
  vpninfo->useragent = openconnect_create_useragent("Open AnyConnect VPN Agent");
  vpninfo->mtu = 1406;
@@ -202,7 +207,7 @@
  else
   vpninfo->localname = "localhost";
 
- while ((opt = getopt_long(argc, argv, "bC:c:Ddg:hi:k:K:lp:Q:qSs:U:u:Vvx:",
+ while ((opt = getopt_long(argc, argv, "bC:c:Ddg:hi:k:K:lpP:Q:qSs:U:u:Vvx:",
       long_options, NULL))) {
   if (opt < 0)
    break;
@@ -298,6 +303,9 @@
   case 'p':
    vpninfo->cert_password = optarg;
    break;
+  case 'P':
+   openconnect_parse_proxyurl(vpninfo, optarg);
+   break;
   case 's':
    vpninfo->vpnc_script = optarg;
    break;
diff -Naur openconnect-2.12/openconnect.h openconnect-2.12.ilap/openconnect.h
--- openconnect-2.12/openconnect.h 2009-12-08 05:40:34.000000000 +1300
+++ openconnect-2.12.ilap/openconnect.h 2009-12-17 22:04:33.000000000 +1300
@@ -44,6 +44,9 @@
 #define OC_FORM_OPT_SELECT 3
 #define OC_FORM_OPT_HIDDEN 4
 
+#define OC_DEF_SSL_PORT  "443"
+#define OC_DEF_PROXY_PORT "8080"
+
 /* char * fields are static (owned by XML parser) and don't need to be
    freed by the form handling code -- except for value, which for TEXT
    and PASSWORD options is allocated by process_form() when
@@ -143,6 +146,10 @@
  char sid_tokencode[9];
  char sid_nexttokencode[9];
 
+ char *proxy;
+ char *proxy_port;
+ char *port;
+
  const char *localname;
  char *hostname;
  char *urlpath;
@@ -319,6 +326,8 @@
 /* http.c */
 int openconnect_obtain_cookie(struct openconnect_info *vpninfo);
 char *openconnect_create_useragent(char *base);
+void openconnect_parse_proxyurl(struct openconnect_info *vpninfo, char *url);
+int process_http_proxy(struct openconnect_info *vpninfo, int ssl_sock);
 
 /* ssl_ui.c */
 int set_openssl_ui(void);
diff -Naur openconnect-2.12/ssl.c openconnect-2.12.ilap/ssl.c
--- openconnect-2.12/ssl.c 2009-12-08 05:40:34.000000000 +1300
+++ openconnect-2.12.ilap/ssl.c 2009-12-18 00:30:04.000000000 +1300
@@ -477,17 +477,23 @@
  int ssl_sock = -1;
  int err;
 
+ char *hostname = vpninfo->hostname;
+ char *port = vpninfo->port;
+
+ if (vpninfo->proxy) {
+  hostname = vpninfo->proxy; 
+  port = vpninfo->proxy_port; 
+ }
+
  if (vpninfo->peer_addr) {
   ssl_sock = socket(vpninfo->peer_addr->sa_family, SOCK_STREAM, IPPROTO_IP);
   if (ssl_sock < 0) {
   reconn_err:
-   vpninfo->progress(vpninfo, PRG_ERR, "Failed to reconnect to host %s\n", vpninfo->hostname);
+   vpninfo->progress(vpninfo, PRG_ERR, "Failed to reconnect to host %s\n", hostname);
    return -EINVAL;
   }
   if (connect(ssl_sock, vpninfo->peer_addr, vpninfo->peer_addrlen))
    goto reconn_err;
-
-  
   
  } else {
   struct addrinfo hints, *result, *rp;
@@ -501,14 +507,15 @@
   hints.ai_addr = NULL;
   hints.ai_next = NULL;
 
-  err = getaddrinfo(vpninfo->hostname, "443", &hints, &result);
+  err = getaddrinfo(hostname, port, &hints, &result);
+
   if (err) {
    vpninfo->progress(vpninfo, PRG_ERR, "getaddrinfo failed: %s\n", gai_strerror(err));
    return -EINVAL;
   }
 
   vpninfo->progress(vpninfo, PRG_INFO,
-      "Attempting to connect to %s\n", vpninfo->hostname);
+      "Attempting to connect to %s\n", hostname);
 
   for (rp = result; rp ; rp = rp->ai_next) {
    ssl_sock = socket(rp->ai_family, rp->ai_socktype,
@@ -534,12 +541,19 @@
   freeaddrinfo(result);
   
   if (ssl_sock < 0) {
-   vpninfo->progress(vpninfo, PRG_ERR, "Failed to connect to host %s\n", vpninfo->hostname);
+   vpninfo->progress(vpninfo, PRG_ERR, "Failed to connect to host %s\n", hostname);
    return -EINVAL;
   }
  }
  fcntl(ssl_sock, F_SETFD, FD_CLOEXEC);
 
+ if (process_http_proxy(vpninfo, ssl_sock) <= 0) {
+  vpninfo->progress(vpninfo, PRG_ERR, "Proxy connection failure\n");
+  SSL_free(https_ssl);
+  close(ssl_sock);
+  return -EINVAL;
+ }
+
  ssl3_method = TLSv1_client_method();
  if (!vpninfo->https_ctx) {
   vpninfo->https_ctx = SSL_CTX_new(ssl3_method);
diff -Naur openconnect-2.12/version.c openconnect-2.12.ilap/version.c
--- openconnect-2.12/version.c 2009-12-08 05:40:34.000000000 +1300
+++ openconnect-2.12.ilap/version.c 2009-12-18 00:30:29.000000000 +1300
@@ -1 +1 @@
-char openconnect_version[] = "v2.12";
+char openconnect_version[] = "v2.12-unknown";

So, the steps to get it work are the following:

#1. Download and unpack the openconnect-2.12 source.
$ cd /path/to/devel/dir
...
$ wget ftp://ftp.infradead.org/pub/openconnect/openconnect-2.12.tar.gz
...
$ tar zxvf openconnect-2.12.tar.gz
...

#2. Save this patch to your box.
$ vi openconnect-2.12_proxy.diff
i
COPY&PASTE the patch above.
ESC
:x!

#3. Patch the source

$ cd openconnect-2.12
$ patch < ../openconnect-2.12_proxy.diff
patching file http.c
patching file main.c
patching file openconnect.h
patching file ssl.c
patching file version.c

$ make
...
...
$ sudo make install
...
...

#4. Test it...
$ sudo openconnect -v --proxy=my.proxy.com target.host.com
Attempting to connect to my.proxy.com
Connecting to the proxy my.proxy.com on port 8080
==>CONNECT target.host.com:443 HTTP/1.1
Host: target.host.com
User-Agent: Open AnyConnect VPN Agent v2.12-unknown
Proxy-Connection: keep-alive
Connection: keep-alive
Accept-Encoding: identity

<==
SSL negotiation with target.host.com
Connected to HTTPS on target.host.com
GET target.host.com/
Got HTTP response: HTTP/1.1 302 Object Moved
Content-Type: text/html; charset=UTF-8
Content-Length: 0
Cache-Control: no-cache
Pragma: no-cache
Connection: Keep-Alive
Date: Thu, 17 Dec 2009 11:31:29 GMT
Location: /+webvpn+/index.html
Set-Cookie: tg=; expires=Thu, 01 Jan 1970 22:00:00 GMT; path=/; secure
GET target.host.com/+webvpn+/index.html
Got HTTP response: HTTP/1.1 200 OK
Server: Virata-EmWeb/R6_2_0
Transfer-Encoding: chunked
Content-Type: text/xml
Cache-Control: max-age=0
Set-Cookie: webvpn=; expires=Thu, 01 Jan 1970 22:00:00 GMT; path=/; secure
Set-Cookie: webvpnc=; expires=Thu, 01 Jan 1970 22:00:00 GMT; path=/; secure
Set-Cookie: webvpnlogin=1; secure
X-Transcend-Version: 1
Fixed options give
Please enter your username and password.
Username:yourusername
Password:yourpassword
POST target.host.com/+webvpn+/index.html
Got HTTP response: HTTP/1.1 200 OK
Server: Virata-EmWeb/R6_2_0
Transfer-Encoding: chunked
Content-Type: text/xml
Cache-Control: max-age=0
Set-Cookie: webvpnlogin=; expires=Thu, 01 Jan 1970 22:00:00 GMT; path=/; secure
Set-Cookie: webvpn=.............; path=/; secure
Set-Cookie: webvpnc=bu:/CACHE/stc/...........
X-Transcend-Version: 1
Got CONNECT response: HTTP/1.1 200 OK
X-CSTP-Version: 1
X-CSTP-Address: 123.123.123.231
X-CSTP-Netmask: 255.255.248.0
X-CSTP-DNS: 222.222.222.228
X-CSTP-DNS: 222.222.222.229
X-CSTP-Lease-Duration: 1209600
X-CSTP-Session-Timeout: none
X-CSTP-Idle-Timeout: 1800
X-CSTP-Disconnected-Timeout: 1800
X-CSTP-Default-Domain: company.com
X-CSTP-Split-Include: 222.222.221.0/255.255.248.0
X-CSTP-Split-Include: 222.222.222.0/255.255.240.0
X-CSTP-Keep: true
X-CSTP-Rekey-Time: 14400
X-CSTP-Rekey-Method: ssl
X-CSTP-DPD: 30
X-CSTP-Keepalive: 20
X-CSTP-Banner: Successfully%20authenticated%21%0AYou%27re%20going%20to%20connect%20a%20SSL%20gateway%20to%20the%20.............
X-CSTP-Smartcard-Removal-Disconnect: true
X-CSTP-Content-Encoding: deflate
X-DTLS-Session-ID: 5820B98BXXXXXXXXXXXXXXA62D88086306
X-DTLS-Port: 443
X-DTLS-Keepalive: 20
X-DTLS-DPD: 30
X-DTLS-Rekey-Time: 14400
X-CSTP-MTU: 1406
X-DTLS-CipherSuite: AES128-SHA
CSTP connected. DPD 30, Keepalive 20
DTLS option X-DTLS-Session-ID : 5820B98BXXXXXXXXXXXXXXA62D88086306
DTLS option X-DTLS-Port : 443
DTLS option X-DTLS-Keepalive : 20
DTLS option X-DTLS-DPD : 30
DTLS option X-DTLS-Rekey-Time : 14400
DTLS option X-DTLS-CipherSuite : AES128-SHA
SSL_set_session() failed with old protocol version 0x100
Your OpenSSL may lack Cisco compatibility support
See http://rt.openssl.org/Ticket/Display.html?id=1751
Use the --no-dtls command line option to avoid this message
Set up DTLS failed; using SSL instead
Connected tun0 as 123.123.123.231, using SSL + deflate
Did no work; sleeping for 20000 ms...
^C
Send BYE packet: Client received SIGINT

Grow /tmp Filesystem (TMPFS) on-line under Solaris 10.

2009-11-29T16:59:00.000-08:00

I was asked recently whether it is possible to grow the /tmp Filesystem on-line (/wo reboot) under Solaris or not. Because the /tmp was configured to 512MB mount and it's exceeded and unfortunately that box couldn't be rebooted nor the /tmp umounted (bunch of files were opened and used). The err msg was similar as these ones below.
... tmpfs: [ID ... kern.warning] WARNING: /tmp: File system full, swap space limit exceeded
or
... tmpfs: [ID ... kern.warning] WARNING: /tmp: File system full, memory allocation failed

Unfortunately, the mount_tmpfs doesn't support the "remount" option ("mount -F tmpfs -o remount /tmp"), so I had to answer "No, it's not".

But, I couldn't help thinking about this. I know that the TMPFS (a.k.a. RAMFS in some other systems) is using the anonymous pages for the TMPFS. So, why the grow of a TMPFS is impossible?

I began to dig a bit in the header files to get some relevant info about the current VFS layer of Solaris.

### 1. I checked the current VFS structure in the /usr/include/sys/vfs.h
(almost every modern OS is using that structure).

As I remebered from the old-days, that the private data should be the pointer to the
per-mount data structure for any FS.


typedef struct vfs {
...
uint_t          vfs_bsize;              /* native block size */
...
void            *vfs_data;              /* private data */
...
} vfs_t

So, I found that the "vfs_data" should includes the per-mount basis info which I was looking for.

###2. /usr/include/sys/fs/tmp.h
I was looking for the header file of the TMPFS which should consists the declaration of this private structure.

/*
* tmpfs per-mount data structure.
...
struct tmount {
...
char            *tm_mntpath;    /* name of tmpfs mount point */
...
ulong_t         tm_anonmax;     /* file system max anon reservation */
pgcnt_t         tm_anonmem;     /* pages of reserved anon memory */
...
};

I found it, but I wasn't sure that this info needs for me or not.
I had to do some more checks that the tm_anonmax or tm_anonmem is using to calculate the size of the FS or not.

I run the following info to get the values of the relevant variables.

@@ 1. Get the size of the /tmp FS.
# grep tmpfs /etc/vfstab *NOTE: Get the mount options.
swap - /tmp tmpfs - yes size=2048m

NOTE: If the "size" option is set, the /tmp will be mounted on that size and if it's
not then it use all of the available virtual memory. See man mount_tmpfs.

# df -h /tmp
Filesystem size used avail capacity Mounted on
swap 2.0G 272K 2.0G 1% /tmp

@@ 2. Get the pointer to the VFS struct of the /tmp FS.
# echo "::fsinfo" | mdb -k | egrep "VFSP|/tmp"
VFSP FS MOUNT
000003000f46d840 tmpfs /tmp

@@ 3. Get the pointer to the private data struct of the /tmp FS and its blocskize.
# echo "3000f46d840::print -ta vfs_t vfs_bsize vfs_data" | mdb -k
3000f46d864 uint_t vfs_bsize = 0x2000 *NOTE 1:0x2000 means 8192 (8KB) in Decimal.
3000f46d878 void *vfs_data = 0x3000f488ce8

*NOTE 2: The vfs_bsize (block size of the FS) is equal /w the pagesize as this FS is
stored in the virtual memory.

@@ 4. Get the values of all of the interested variables.
# echo "3000f46d840::print vfs_t vfs_data | ::print struct tmount tm_mntpath tm_anonmax tm_anonmem" | mdb -k
tm_mntpath = 0x3000f5f9508 "/tmp"
tm_anonmax = 0x40000 *NOTE 2: 0x40000 means 256KB
tm_anonmem = 0x22

As you can see the anonmax's (Nr. of pages) 0x40000 * pagesize (0x2000) = 0x80000000 = 2GB.
So, probably the tm_anonmax is the relevant size info of the /tmp FS.

But, I wasn't sure. So, what else could have done to be sure that's the relevant variable to /tmp size.
Yep, the Opensolaris source tree. So, to be sure that the size of the /tmp is calculated from the
tm_anonmax and vfs_bsize (a.k.a pagesize), I had to dig a bit in the OpenSolaris source.

Long story sort, I found an evidence that the tm_anonmax is used to set the size of the /tmp FS.
See, brief details below (Source code of the tmp_vfsops.c):

...
static int
230 tmp_mount(
...
236     struct tmount *tm = NULL; // NOTE 3: tm is a pointer to a tmount structure

...

276      * tm_anonmax is set according to the mount arguments
277      * if any.  Otherwise, it is set to a maximum value.
278      */
279     if (vfs_optionisset(vfsp, "size", &sizestr)) { // NOTE: if the size vfstab option is set, SEE  above
280         if ((error = tmp_convnum(sizestr, &anonmax)) != 0)
281             goto out;
282     } else {
283         anonmax = ULONG_MAX; // NOTE: If the size is not set then use all...
284     }
...
314     tm->tm_anonmax = anonmax; // NOTE:set the Private data's variable
315
316     vfsp->vfs_data = (caddr_t)tm; //NOTE: Private data is set.
...
319     vfsp->vfs_bsize = PAGESIZE; // Yep, bsize is equal PAGESIZE.

A live example.

This example shows how I grew the 2GB mounted /tmp to 3GB on a Solaris 10 5/09 u7 SPARC system. Please, don't try this in your system.

If you don't listen to me then keep in mind that the HEXA numbers might differ on your system.

#1. Get the relevant info of the /tmp

# df -h /tmp
Filesystem size used avail capacity Mounted on
swap 2.0G 272K 2.0G 1% /tmp

# echo "::fsinfo" | mdb -k | egrep "VFSP|/tmp"
VFSP FS MOUNT
000003000f46d840 tmpfs /tmp

# Get the address of the tm_anonmax to set its value.
# echo "000003000f46d840::print vfs_t vfs_data | ::print -ta struct tmount tm_anonmax" | mdb -k
3000f488d00 tm_anonmax = 0x40000
// Address and the currrent value of the tm_anonmax

### 2. Set the new value
# echo "3000f488d00/Z 0x60000" | mdb -kw
0x3000f488d00: 0x40000 = 0x60000
NOTE: the 0x60000 is 384KB -> 384KB * 8KB = 3072KB = 3GB.

### 3. Check if it's set.
# echo "3000f488d00/J" | mdb -k
0x3000f488d00: 60000
OR
# echo "3000f46d840::print vfs_t vfs_data | ::print struct tmount tm_anonmax" | mdb -k
tm_anonmax = 0x60000

### 4. Check if it's working or not.
# df -h /tmp
Filesystem size used avail capacity Mounted on
swap 3.0G 272K 3.0G 1% /tmp

Yep, it done...

Conclusion:
The Live Example above shows a real example which was done by myself on a Sol10 system.
I think this solution above can be done on any Solaris version which has the feature of the tmpfs-ed /tmp Filesystem. But, you should use some different method (adb) to get same kind of info in different version of Solaris.

This was my 3rd non-existent-solution-before-this type of post.:)
This post or any info in this post is not a copy from any other post.
So, there wasn't any solution about this on the Internet until this post.
Or maybe was, but I couldn't find any.:)

Unfortunately, it can't be done in a zone as the zone doesn't allow you to access to the /dev/kmem and /dev/ksyms. You, should access them from the Global Zone, BUT I haven't tried...

Disclaimer:
First, I take no responsibility for any trouble caused by this info and make no guarantees of any kind.
So, please be very carefull when you want use any info from this post.
SO, YOU'RE USING ANY INFO IN THIS POST AT YOUR OWN RISK!

Cisco VPN Client on Karmic Koala

2009-09-06T02:53:00.000-07:00

=============================================
UPDATE:
=============================================
Issue: I've got a new Intel Core 2 Duo based Notebook and after some minutes the system's hang (black screen and the mouse's not responding).

Cause: The kernel's paniced due the ipsec's kernel module. I switched to a virtual console (CTRL+ALT+F2) as soon as I started the vpnclient to get any error message from the system when the system hung as I didn't find any relevant error message in any log files.
After some seconds/minutes later, I saw the panic message on the virtual console, which was relating to the cisco_ipsec.ko... So, the panic was related to the Cisco VPN kernel module.

Workaround: Probably, you're using an SMP based kernel on an SMP capable arch (For ex. Intel Core 2 Duo). You should disable a CPU while you're using the VPN Cisco client.
See, this comment to solve.
=============================================

I've recently installed Kubuntu 9.10 (Karmic Koala) Alpha 5 on my home laptop (IBM ThinkPad T60).
It comes with a lot of nice features like support for the Cisco VPN ("network-manager-vpnc" a.k.a. VPNC plugin) and so on, but unfortunately the IPSec over TCP (TCP tunnel), which method needs to be used at my company, is not supported by the VPNC.
So, I must have tried to install the latest Cisco VPN Client (4.8.02.0300) on my fresh install.

Steps:
### 1. Get the latest vpnclient from the Net
Google it

### 2. Uncompress and compile
$ tar zxf vpnclient-linux-x86_64-4.8.02.0030-k9.tar.gz && cd vpnclient
$ sudo ./vpn_install
Cisco Systems VPN Client Version 4.8.02 (0030) Linux Installer
...
Making module
make -C /lib/modules/2.6.31-9-generic/build SUBDIRS=/opt/Devel/CiscoVPNClient/4.8.02/vpnclient modules
make[1]: Entering directory `/usr/src/linux-headers-2.6.31-9-generic'
CC [M] /opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.o
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c: In function ‘interceptor_init’:
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c:132: error: ‘struct net_device’ has no member named ‘hard_start_xmit’
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c:133: error: ‘struct net_device’ has no member named ‘get_stats’
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c:134: error: ‘struct net_device’ has no member named ‘do_ioctl’
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c: In function ‘add_netdev’:
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c:271: error: ‘struct net_device’ has no member named ‘hard_start_xmit’
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c:272: error: ‘struct net_device’ has no member named ‘hard_start_xmit’
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c: In function ‘remove_netdev’:
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.c:294: error: ‘struct net_device’ has no member named ‘hard_start_xmit’
make[2]: *** [/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.o] Error 1
make[1]: *** [_module_/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori] Error 2
make[1]: Leaving directory `/usr/src/linux-headers-2.6.31-9-generic'
make: *** [default] Error 2
Failed to make module "cisco_ipsec.ko".

Oooops, so it's failed. I tried to find any solution on the Net but I couldn't get any.
I looked at the source code and seemed that the Linux kernel 2.6.31 replaced the old net_device structure with a new one and is using that new structure for the net_device operations. This structure called net_device_ops (const struct net_device_ops *netdev_ops).

I think this new netdevice feature is available from the 2.6.29 kernel and it was coexisted with the old method too.
I got some info about this on the following link: http://cateee.net/lkddb/web-lkddb/COMPAT_NET_DEV_OPS.html

Unfortunately, this option is disappeared from the 2.6.31 so there was no option anymore to get the vpnclient to work with a COMPAT_NET_DEV_OPS=y built 2.6.31 kernel.

As I didn't find any info how should I get the Cisco vpnclient to work on the Linux kernel 2.6.31 I needed to try to modify the Cisco's source(s) myself. Long story short I finally managed to create a small patch which is working for me on the Kubuntu 9.10 Karmic Koala /w Linux kernel 2.6.31-9.

Here comes the tiny patch.

$ cat vpnclient-linux.2.6.31.diff


--- vpnclient.ori/interceptor.c 2009-05-21 01:16:34.000000000 +1200
+++ vpnclient/interceptor.c 2009-09-06 22:02:39.000000000 +1200
@@ -116,6 +116,14 @@
};
#endif

+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,31)
+static struct net_device_ops interceptor_netdev_ops = {
+    .ndo_start_xmit = interceptor_tx,
+    .ndo_do_ioctl   = interceptor_ioctl,
+    .ndo_get_stats  = interceptor_stats,
+};
+#endif
+
static struct notifier_block interceptor_notifier = {
.notifier_call = handle_netdev_event,
};
@@ -129,9 +137,13 @@
{
ether_setup(dev);

+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,31)
+    dev->netdev_ops = &interceptor_netdev_ops;
+#else
dev->hard_start_xmit = interceptor_tx;
dev->get_stats = interceptor_stats;
dev->do_ioctl = interceptor_ioctl;
+#endif

dev->mtu = ETH_DATA_LEN-MTU_REDUCTION;
kernel_memcpy(dev->dev_addr, interceptor_eth_addr,ETH_ALEN);
@@ -242,6 +254,9 @@
{
int rc = -1;
int i = 0;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,31)
+    struct net_device_ops * tmp_ops;
+#endif

if (!supported_device(dev))
{
@@ -268,8 +283,14 @@
Bindings[i].original_mtu = dev->mtu;

/*replace the original send function with our send function */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,31)
+    Bindings[i].InjectSend = dev->netdev_ops->ndo_start_xmit;
+    tmp_ops = (struct net_device_ops *) dev->netdev_ops;
+    tmp_ops->ndo_start_xmit = replacement_dev_xmit;
+#else
Bindings[i].InjectSend = dev->hard_start_xmit;
dev->hard_start_xmit = replacement_dev_xmit;
+#endif

/*copy in the ip packet handler function and packet type struct */
Bindings[i].InjectReceive = original_ip_handler.orig_handler_func;
@@ -285,13 +306,21 @@
{
int rc = -1;
BINDING *b;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,31)
+    struct net_device_ops * tmp_ops;
+#endif

b = getbindingbydev(dev);

if (b)
{
   rc = 0;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,31)
+        tmp_ops = (struct net_device_ops *) dev->netdev_ops;
+        tmp_ops->ndo_start_xmit = b->InjectSend;
+#else
   dev->hard_start_xmit = b->InjectSend;
+#endif
   kernel_memset(b, 0, sizeof(BINDING));
}
else

So, the further steps to get cisco vpnclient to work:

### 4. Save this small patch above to a file. for example to cisco-vpnclient.2.6.31.diff and do the folowing:
Or downolad it from HERE

### 5. Patch the source
$ cd vpnclient
$ patch < ../vpnclient-linux.2.6.31.diff
patching file interceptor.c

### 6. Compile it
$ sudo ./vpn_install
Cisco Systems VPN Client Version 4.8.02 (0030) Linux Installer
Copyright (C) 1998-2006 Cisco Systems, Inc. All Rights Reserved.
...
* Modules will be installed in "/lib/modules/2.6.31-9-generic/CiscoVPN".
...
Making module
make -C /lib/modules/2.6.31-9-generic/build SUBDIRS=/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori modules
make[1]: Entering directory `/usr/src/linux-headers-2.6.31-9-generic'
CC [M] /opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/interceptor.o
CC [M] /opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/linuxkernelapi.o
LD [M] /opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/cisco_ipsec.o
Building modules, stage 2.
MODPOST 1 modules
WARNING: could not find /opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/.libdriver.so.cmd for
/opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/libdriver.so
CC /opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/cisco_ipsec.mod.o
LD [M] /opt/Devel/CiscoVPNClient/4.8.02/vpnclient.ori/cisco_ipsec.ko
make[1]: Leaving directory `/usr/src/linux-headers-2.6.31-9-generic'
Copying module to directory "/lib/modules/2.6.31-9-generic/CiscoVPN".
...

### 7. Load the compiled kernel module (cisco_ipsec.ko)
$ lsmod | grep cisco
$ sudo /etc/init.d/vpnclient_init start
Starting /opt/cisco-vpnclient/bin/vpnclient: Done
$ lsmod | grep cisco
cisco_ipsec 595804 0

### 8. Connect
$ sudo vpnclient connect WORK
Cisco Systems VPN Client Version 4.8.01 (0640)
Copyright (C) 1998-2007 Cisco Systems, Inc. All Rights Reserved.
Client Type(s): Linux
Running on: Linux 2.6.31-9-generic #29-Ubuntu SMP Sun Aug 30 17:39:23 UTC 2009 i686
Config file directory: /etc/opt/cisco-vpnclient

Initializing the VPN connection.
Initiating TCP to , port
Contacting the gateway at
User Authentication for WORK...

Enter Username and Password.

Username []:
Password []:
Authenticating user.
Negotiating security policies.
Securing communication channel.

"Use of this network is restricted to authorized users only. User activity may be monitored and/or recorded. Anyone using this network expressly consents to such monitoring and/or recording. BE ADVISED: if possible criminal activity is detected, these records, along with certain personal information, may be provided to law enforcement officials."
Do you wish to continue? (y/n): y

Your VPN connection is secure.

VPN tunnel information.
Client address:
Server address:
Encryption: 128-bit AES
Authentication: HMAC-SHA
IP Compression: None
NAT passthrough is active on port TCP
Local LAN Access is enabled

Finally, I get it to work and it seems to be stable as I couldn't experienced any issue with this solution since it's applied.

Reference(s):
http://projects.tuxx-home.at/?id=cisco_vpn_client

My 1st HTPC #Part 3: Mini PCI Capture Card - AverMedia M103 (hybrid)

2009-04-05T03:44:00.000-07:00

This part sometimes called "AverTV Hybrid DVB-T" or "Avermedia M103 (hybrid)" or whatever, but the most important things is which chips are built into this card and are those supported by the Linux or not.

XC3028 - Xceive Silicon Tuner (for Hybrid system)
MT352 - Silicon Tuner - Zarlink DVB-T Demodulator (Digital Terrestrial TV Demodulator)
SAA7135H - Philips Video Decoder and Broadcast Audio Decoder / PCI bridge (for worldwide sku)

Fortunately, these chips are (theoreticaly) supported by the latest Linux kernel >=2.6.xx, so let's to see that the card can work on it or not.

This card has just one input and doesn't any external output (Audio, S-video, composit and A/V output).
The input of this card is the A/V output of my SKY Decoder (Pace DS620) and the output is the Analog VGA which will be an input for a VGA port of my Samsung LE32R27B TV.

As this card doesn't have any external audio output so the Audio just could come from the card directly by the DMA using the PCI bus.
I've just created a semantics picture of the capture card and its relations to other components to make sense how it works. See above.

root@htpd:/# dmesg | more
...
Linux video capture interface: v2.00
saa7130/34: v4l2 driver version 0.2.14 loaded
saa7134 0000:04:07.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17
saa7133[0]: found at 0000:04:07.0, rev: 209, irq: 17, latency: 0, mmio: 0xfd8fe000
saa7134 0000:04:07.0: setting latency timer to 64
saa7133[0]: subsystem: 1461:f636, board: AVerMedia MiniPCI DVB-T Hybrid M103
[card=145,autodetected]
saa7133[0]: board init: gpio is 360000
saa7133[0]: i2c eeprom 00: 61 14 36 f6 00 00 00 00 00 00 00 00 00 00 00 00
...
saa7133[0]: i2c eeprom f0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
saa7133[0]: i2c scan: found device @ 0x1e [???]
saa7133[0]: i2c scan: found device @ 0xa0 [eeprom]
saa7133[0]: i2c scan: found device @ 0xc2 [???]
tuner' 0-0061: chip found @ 0xc2 (saa7133[0])
xc2028 0-0061: creating new instance
xc2028 0-0061: type set to XCeive xc2028/xc3028 tuner
firmware: requesting xc3028-v27.fw
xc2028 0-0061: Loading 80 firmware images from xc3028-v27.fw, type: xc2028 firmware, ver 2.7
xc2028 0-0061: Loading firmware for type=BASE F8MHZ (3), id 0000000000000000. (0), id 00000000000000ff:
xc2028 0-0061: Loading firmware for type=(0), id 0000000100000007.
xc2028 0-0061: Loading SCODE for type=MONO SCODE HAS_IF_5320 (60008000), id 0000000f00000007.
saa7133[0]: registered device video0 [v4l2]
saa7133[0]: registered device vbi0
saa7134 ALSA driver for DMA sound loaded
saa7133[0]/alsa: saa7133[0] at 0xfd8fe000 irq 17 registered as card -2
xc2028 0-0061: attaching existing instance
xc2028 0-0061: type set to XCeive xc2028/xc3028 tuner
DVB: registering new adapter (saa7133[0])
DVB: registering frontend 0 (Zarlink MT352 DVB-T)...
xc2028 0-0061: Loading firmware for type=(0), id 0000000100000007.
xc2028 0-0061: Loading SCODE for type=MONO SCODE HAS_IF_5320 (60008000), id 0000000f00000007.
...
So, it can be seen that all of the chips are recognized and configured by the intrepid and also that the following devices are registered
saa7133[0]: registered device video0 [v4l2] as /dev/video0 (as there is no other capture card)
saa7133[0]: registered device vbi0 as /dev/vbi0 (data send within the channel such the teletext and so on)
saa7134 ALSA driver for DMA sound loaded as /dev/dsp1 as the intel HDA is using the /dev/dsp audio device.
DVB: registering frontend 0 (Zarlink MT352 DVB-T)... as /dev/dvb/adapter0/*

Issues and resolutions
OK, let's check that it works or not. I have some options to check this but I prefer the mplayer or the tvtime and using the aerial antenna or the output of my SKY box. So, to fully understand what I wanted to check see the picture below.

As my htpc is connected to an old pace DSR600 Sky decoder which is using the UHF 38th channel RF output (607.250 MHz in Europian frequency table) I needed to check that frequency with any TV capable player/application can get the TV stream or not. See details: tvtime, mplayer, UHF, VHF, Channel_37.

Also keep in mind that the "AVerMedia MiniPCI DVB-T Hybrid" doesn't have any external audio output which could have been patched to the 6-jack, 2-jack or any TV audio input to get any sound.
In this case the audio output of the capture card must be piped to the Intel HDA output to get any sound. The mplayer is capable to do this but not the tvtime.
To test that capture card I needed to do the following:

1. Get which Capture (Audio input) and Playback (Audio output) devices are exist in the system.
ilap@htpc:~$ arecord -l
**** List of CAPTURE Hardware Devices ****
card 0: Intel [HDA Intel], device 0: ALC888 Analog [ALC888 Analog]
Subdevices: 1/1
Subdevice #0: subdevice #0 card 0: Intel [HDA Intel], device 2: ALC888 Analog [ALC888 Analog]
Subdevices: 1/1
Subdevice #0: subdevice #0 card 1: SAA7134 [SAA7134], device 0: SAA7134 PCM [SAA7134 PCM]
Subdevices: 1/1 Subdevice #0: subdevice #0

ilap@htpc:~$ aplay -l
**** List of PLAYBACK Hardware Devices ****
card 0: Intel [HDA Intel], device 0: ALC888 Analog [ALC888 Analog]
Subdevices: 1/1
Subdevice #0: subdevice #0 card 0: Intel [HDA Intel], device 1: ALC888 Digital [ALC888 Digital]
Subdevices: 1/1 Subdevice #0: subdevice #0

So, it shows 2 Sound cards:

Intel HDA (the internal one) and the
SAA7134 PCM (which is bundled in the Avermedia M103 capture card).

It also clearly seen that the SAA7134 PCM is just capable for the CAPTURE and not for the PLAYBACK as there is no any Audio output in that card.

2. Check the card /w mplayer and/or tvtime
ilap@htpc:~$ mplayer -v tv:// -tv freq=607.250:\ device=/dev/video0:input=0:norm=PAL:\ audiorate=32000:alsa:forceaudio:amode=1:\ adevice=hw.1,0:volume=100:immediatemode=0 -vo gl2 -aspect 16:9 ...

The freq=607.250 (or channel=38), adevice=hw,1.0 and the immediatemode=0 are the most important parameters for the mplayer. The adevice tells which is the audio capture device where the sound comes from. The immediatemode=0 lets the mplayer know that the AUDIO is not patched by cable (capture card's external audio output<--cable to-->other soundcard external input (such mic)) but piped directly from adevice to the default audio output device in this case to /dev/dsp (Intel HDA).

As the tvtime is not capable to do this pipe we need to pipe it directly. For example the following command starts the tvtime in background and the "arecord | aplay" pipe is capturing the audio from the hw:1,0 (sa7134 capture card) and pipe to the default sound device (Intel HDA's /dev/dsp).

ilap@htpc:~$ tvtime & arecord -D hw:1,0 -r 3200 | aplay
...

Finaly, I got the Video and Audio stream so I just needed to configure it in the MythTV. See, later

Conclusion

I was supprised how good quality TV stream is coming out from this card as I was expected a "reasonable" output as it's a nearly zero priced capture card.

But, to tell the truth I compared the output from my htpc and directly attached SKY decoder (through EuroSCART) and the quality of the stream from the directly attached input is better then the htpc one.
But, I think the problem is that the TV has some internal features (such noise reduction etc.) to improve the quality of a noisy channel such a BNC input and I think that/those feature(s) is/are not used for the VGA output. But, I'm maybe wrong as I am/was not able to check the VGA output as an TV input to the TV as I don't have any VGA to TV (RCA/BNC/Composit/EuroSCART) converter.

Anyway it's more than worth buying this capture card for anybody who needs a MiniPCi capture card.

References

http://www.linuxtv.org/wiki/index.php/AverMedia_M103_(Hybrid)_Mini_PCI

My 1st HTPC #Part 2: Mini PCIe Wi-Fi Card - Atheros AR2425 AR50007EG 802.11g

2009-04-05T03:42:00.001-07:00

This card includes an Atheros AR2425 chip and is using the PCI id 186c:001c (Atheros Communications Inc. AR242x 802.11bg Wireless PCI Express Adapter (rev 01)) and Subsystem id 144f:7131 (Askey Computer Corp. Device)

In spite of this chip is theoretically supported by the Linux through the madwifi and/or ath5k drivers I got a big headache to get this card works due the lack of information on the Internet.

See details (issues and resolutions) below...

Issues and resolutions

1. The version of the HAL of the current Madwifi driver (0.9.4) in Mythbuntu 8.10 doesn't support this card.
Error message:
ath_hal: module license 'Proprietary' taints kernel.
ath_hal: 0.9.30.13 (AR5210, AR5211, AR5212, AR5416, RF5111, RF5112, RF2413, RF5413, RF2133) wlan: 0.8.4.2 (svn r2619)
ath_pci: 0.9.4.5 (svn r2619) PCI: Enabling device 0000:04:00.0 (0000 -> 0002) ACPI: PCI Interrupt 0000:04:00.0[A] -> GSI 19 (level, low) -> IRQ 19
PCI: Setting latency timer of device 0000:04:00.0 to 64
MadWifi: unable to attach hardware: 'Hardware revision not supported' (HAL status 13)
ACPI: PCI interrupt for device 0000:04:00.0 disabled

2. I had to install the latest madwifi. Check the following ticket: http://madwifi-project.org/ticket/1192 to process. The install process of the newest madwifi will ask to remove the current ath_hal and ath_pci drivers. So, I recommend to backup them before remove.

Brief steps to install the latest HAL (as in 03/27/2009):
2.1. Get the sources
# sudo apt-get install svn build-essential linux-headers-$(uname -r)
# cd ~
# svn checkout http://svn.madwifi-project.org/madwifi/branches/madwifi-hal-0.10.5.6 madwifi-hal-0.10.5.6

2.2. Unload the already loaded driver (if they're not already blacklisted)
# cd madwifi-hal-0.10.5.6/scripts # sudo ./madwifi-unload
# sudo ./find-madwifi-modules.sh $(uname -r)
2.3. Build the driver
# cd .. # make && sudo make install # sudo modprobe ath_pci
But, the latest madwifi (0.10.9.6) couldn't configure the Mini-PCIe card.
Error message:
wifi%d: unable to attach hardware: ‘EEPROM mac address invalid’ (HAL status 9)
I should have used the ath5k instead.

4. I tried to use the ath5k driver instead, but it is not built for the Ubuntu 8.10. So, I needed to install the backported ath5k driver.

Brief steps to install
4.1. Install the backports
# sudo apt-get install linux-backports-modules-generic
# sudo (/bin/echo -e "blacklist ath_hal\nblacklist ath_pci" >> /etc/modprobe.d/blacklisted)
# sudo init 6

Unfortunately, the backported ath5k still couldn't configure the Mini-PCIe card.
Err msg:
cfg80211: Using static regulatory domain info
cfg80211: Regulatory domain: US
(start_freq - end_freq @ bandwidth), (max_antenna_gain, max_eirp)
(2402000 KHz - 2472000 KHz @ 40000 KHz), (600 mBi, 2700 mBm)
(5170000 KHz - 5190000 KHz @ 40000 KHz), (600 mBi, 2300 mBm)
(5190000 KHz - 5210000 KHz @ 40000 KHz), (600 mBi, 2300 mBm)
(5210000 KHz - 5230000 KHz @ 40000 KHz), (600 mBi, 2300 mBm)
(5230000 KHz - 5330000 KHz @ 40000 KHz), (600 mBi, 2300 mBm)
(5735000 KHz - 5835000 KHz @ 40000 KHz), (600 mBi, 3000 mBm)
cfg80211: Calling CRDA for country: US
ath5k 0000:02:00.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17
ath5k 0000:02:00.0: registered as 'phy0'
ath5k 0000:02:00.0: PCI INT A disabled
ath5k: probe of 0000:02:00.0 failed with error -22 #The -22 is the EINVAL from errno.h

So, none of the supported drivers (madwifi, ath5k, ath9k) could configure it at all.
In case of I never give the things like this up, I was thinking of using the ndiswrapper as the last chance.
But, before the last chance I had a quick rest /w a cup of coffee and was thinking of the things of the BIG LIFE:)

After these, I installed the ndiswrapper and dowloaded the Windows native driver to use by the ndiswrapper.

4. Install the ndiswrapper and blacklist the ath5k too.
# sudo (/usr/bin/echo "blacklist ath5k" >> /etc/modprobe.d/blacklisted)
# rmmod ath5k
# sudo apt-get install ndisgtk
# cd ~
# wget http://blakecmartin.googlepages.com/ar5007eg-64-0.2.tar.gz
# tar xvf ar5007eg-*.tar.gz # sudo ndiswrapper -i net5211.inf
# sudo init 6
# modprobe ndiswrapper
# ifconfig wlan0
wlan0 Link encap:Ethernet HWaddr FF:FF:FF:FF:FF:FF BROADCAST RUNNING MTU:1500 Metric:1 ...

Fortunately, the card has been partialy configured as the native driver was loaded and the interface has been created (wlan0), but the interface still didn't come up as the MAC address was a broadcast one (FF:FF:FF:FF:FF:FF).

So, I tried to modify the MAC with the macchanger, which was successfull, and the response of the 1st ping from my sweety WL-500gP has been arrived healthy.

# sudo apt-get install macchanger
# lspci -vvnn -s 02:0.0
02:00.0 Ethernet controller [0200]: Atheros Communications Inc. AR242x 802.11abg Wireless PCI Express Adapter [168c:001c] (rev 01) Subsystem: Askey Computer Corp. Device [144f:7131]

# macchanger --list=Askey
Misc MACs: Num MAC Vendor --- --- ------
6068 - 00:90:96 - Askey Computer Corp. Wireless MACs: Num MAC Vendor --- --- ------

# macchanger -m 00:90:96:4f:5c:e2 wlan0
# ping cerberus PING cerberus (192.168.0.1) 56(84) bytes of data. 64 bytes from cerberus (192.168.0.1): icmp_seq=1 ttl=64 time=1.32 ms ...

So, after this, I stored the new MAC permanently into the EEPROM as the macchanger just changing it temporary in the kernel space. To do that I used the ath_info command.

# mkdir ath_info && cd ath_info
# svn checkout http://svn.madwifi-project.org/ath_info/trunk/ .
# make
gcc -g -O2 -W -Wall -c ath_info.c
ath_info.c: In function ‘main’:
ath_info.c:2847: warning: ignoring return value of ‘fwrite’, declared with attribute warn_unused_result
gcc -g -O2 -W -Wall -o ath_info ath_info.o

To get which memory address can be used to program the EEPROM
# lspci -v -s 02:0.0 | grep Mem
Memory at fddf0000 (64-bit, non-prefetchable) [size=64K]
Dump the EEPROM addresses where the MAC is stored (0x1d-0x1f and 0xa5-0xa7)

# sudo ./ath_info -d fddf0000 | egrep "^00(18|a0):"
0018: 0000 0000 0000 0000 0000 0000 0000 0000
00a0: 0000 ffff ffff ffff ffff ffff ffff ffff

The MAC (for example: AA:BB:CC:DD:EE:FF) is stored in the following format:
---- 0x18 0x19 0x1a 0x1b 0x1c 0x1d 0x1e 0x1f
0018: 0000 0000 0000 0000 0000 EEFF CCDD AABB
---- 0xa0 0xa1 0xa2 0xa3 0xa4 0xa5 0xa6 0xa7
00a0: 0000 ffff ffff ffff ffff BBAA DDCC FFEE

As the EEPROM of this Mini PCIe card is write protected I must have set some GPIO pin to Low to enable it for writing.
But, which GPIO pin should be set as every motherboard and/or card has different GPIO pin for this?

So, I tried 3 4 5 (the commonly used ones), but no success,
It's come back /w the following error message:

# sudo -w -g 3:0 fddf0000 0x1d 0x5ce2
...
WARNING: The write function may easy brick your device or
violate state regulation on frequency usage.
Proceed on your own risk!
Shall I write the above value(s)? (y/n) y
writing *0x001f := 0x0001
#ERR do_write_pairs: offset 0x001f: wrote 0x5ce2 but read 0x0000
restoring GPIO CR c0 -> 0
...

So, I tried it from 0 and then 1 and the 1 was the winner.
So, the chosen MAC (00:00:e2:4f:5c:e2) is stored with the following commands:

root@htpc:/# sudo -w -g 1:0 fddf0000 0x1d 0x5ce2
...
root@htpc:/# sudo -w -g 1:0 fddf0000 0x1e 0xe24f
...
root@htpc:/# sudo -w -g 1:0 fddf0000 0x1f 0x0000 # This one is not necessary...
...
root@htpc:/# sudo -w -g 1:0 fddf0000 0xa5 0x0000
...
root@htpc:/# sudo -w -g 1:0 fddf0000 0xa6 0x4fe2
...
root@htpc:/# sudo -w -g 1:0 fddf0000 0xa7 0xe25c
root@htpc:/# sudo ./ath_info -d fddf0000 | egrep "^00(18|a0):"
0018: 0000 0000 0000 0000 0000 5ce2 e24f 0000
00a0: 0000 ffff ffff ffff ffff 0000 4fe2 e25c

After this, I disabled the blacklisted drivers (ath_pci, ath_hal) and rebooted the box.
As I expected it's started to do its job.:)

root@htpc:/# dmesg | egrep -i "wifi|ath"
[ 17.796476] ath_hal: module license 'Proprietary' taints kernel.
[ 17.999420] ath_pci 0000:02:00.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17
[ 17.999442] ath_pci 0000:02:00.0: setting latency timer to 64
[ 18.499991] MadWifi: ath_attach: Switching rfkill capability off.
[ 18.667946] wifi0: Atheros AR2425 chip found (MAC 14.2, PHY SChip 7.0, Radio 10.2)
[ 18.808582] ath_pci: wifi0: Atheros 5424/2424: mem=0xfddf0000, irq=17
[ 30.604452] ath0: no IPv6 routers present

# iwconfig ath0
ath0 IEEE 802.11g ESSID:"ilap-wlan" Nickname:""
Mode:Managed Frequency:2.437 GHz Access Point: 00:17:31:FB:2E:44
Bit Rate:11 Mb/s Tx-Power:18 dBm Sensitivity=1/1
Retry:off RTS thr:off Fragment thr:off
Power Management:off
...
# ifconfig ath0
ath0 Link encap:Ethernet HWaddr 00:00:e2:4f:5c:e2
inet addr:192.168.0.3 Bcast:192.168.0.255 Mask:255.255.255.0
inet6 addr: fe80::200:e2ff:fe4f:5ce2/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:452180 errors:0 dropped:0 overruns:0 frame:0
TX packets:143741 errors:0 dropped:0 overruns:0 carrier:0
...

So, the Wi-Fi has been installed...

Diversity mode
Almost every Mini PCI Wi-Fi 802.11b/g card has at least two antennas (MAIN, AUX) and
these two antennas are configured to use the diversity mode by default, which means that the chip dynamically switches between these two antennas to pickup the strongest signal. It sometimes can cause some performance degradation as it's always switching between them.
I just have one antenna, so this setting is not necessary.

# 1. Check the mode
root@htpc:/# grep -r . /proc/sys/dev/wifi0/ 2>/dev/null| egrep "anten|diver"
/proc/sys/dev/wifi0/txantenna:1
/proc/sys/dev/wifi0/rxantenna:0
/proc/sys/dev/wifi0/diversity:1

# 2. Disable the diversity mode
root@htpc:/# echo 0 > /proc/sys/dev/wifi0/diversity
root@htpc:/# echo 2 | tee/proc/sys/dev/wifi0/{r,t}xantenna
or
root@htpc:/# sysctl -w dev.wifi0.diversity=0
root@htpc:/# sysctl -w dev.wifi0.txantenna=2
root@htpc:/# sysctl -w dev.wifi0.rxantenna=2

root@htpc:/# athstats
1 beacon miss interrupts
715 tx management frames
110 tx failed due to too many retries
2112 long on-chip tx retries
573 tx frames with no ack marked
173 tx frames with an alternate rate
3391 rx failed due to bad CRC
401 PHY errors
14 OFDM restart
387 CCK restart
109 periodic calibrations
rssi of last ack: 22
rssi of last rcv: 21
6 switched default/rx antenna
Antenna profile:
[1] tx 42 rx 749
[2] tx 19449 rx 64576

root@htpc:/# 80211stats
19825 rx from wrong bssid
8107 rx discard due to it's a dup
43 rx discard due to mcast echo
17614 rx element unknown
266 rx frame chan mismatch
3 rx w/ incorrect keyid
12 rx seq# violation (CCMP)
29 active scans started
1 nodes timed out inactivity

# To make changes permanent
root@htpc:/# sysctl dev.wifi0 2>/dev/null | egrep "anten|divers"
dev.wifi0.txantenna = 2
dev.wifi0.rxantenna = 2
dev.wifi0.diversity = 0

root@htpc:/sys# echo "# Set the WIFI
dev.wifi0.diversity = 0
dev.wifi0.txantenna = 2
dev.wifi0.rxantenna = 2" >> /etc/sysctl.conf

Conclusion
In spite of the lack of information of this card and hard to get it up, I'm statisfied with it.
It's very cheap, but very good quality. This card is also used by Apple too named it as "Airport Extreme".

References

http://www.madwifi-project.org

My 1st HTPC #Part 1: The hardware

2009-04-04T00:24:00.000-07:00

I have been thinking of building a cheap HTPC (Home Theatre PC) for ages, but when I tried to collect which parts (Barebone Case, CPU, RAM, HD, DVD-ROM, TV-tuner, Graphics card, IR remote, etc.) should I need and how much do those parts cost, the price has reached the 1000 NZD which is much more higher than I should pay for it. So, I thought it is not worth building one...

Also, I thought that a reasonable HW for an HTPC should includes a Core Duo processor with 1GB ram, Wi-Fi, fast HDD, ODD (Optical Disc Drive), TV Capture card, Bluetooth, reasonable Graphic card (just for playback and not for Game). Also, it should be small, quite and flossy.

But, once upon a day:), I saw an Acer Aspire L320 on the TradeMe, here in New Zealand, just for 370 NZD (188 USD) and it includes the Intel(R) ViiV(TM) Technology Platform featuring:

Intel Core Duo processors
Intel i946GZ (North Bridge) + Intel ICH7-DH (South Bridge) chipset
Intel 82573L (Vidalia) Intel PRO/10/100/1000 PM Network Connection.

L320

The Aspire L320 is small (25x20x6cm), quite (~26dB), fast enough and also has some empty Mini-PCI(e) slots (1xMini-PCI and 1x Mini-PCIe) for the additional parts.
So, I thought why not... Let's have a try... If I fail then I will have a slimline PC instead.:)

Anyway, for the Geek I have written the details and put some pictures of the Acer Aspire L320

Case: simline PC
Dimension (WxHxD): 60 x 200 x 250 mm
Weight: 2.6Kg
CD/DVD-ROM: Slim-line Pioneer DVD-RW (DVR-K06RS, 2MB buffer, FW 1.06)
I/O ports:
No any PCI nor ISA expansion slot.

Front Panel o 2-jack Audio Input/output

1x Mini FireWire
Multi-card reader (MMC, SD, MS, MS Pro, XD)
2 USB Slots

Back panel

1xDC input (135W), so I had to choose the low-power components to fit to this restriction.
1xAnalog VGA and 1xDVI port
1xEthernet interface, RJ45 port
4xUSB2.0 port
1xFirewire port
6-jack Audio ports (5.1 but 7.1 /w the front ones)

Downside
There is no

any PCI slot (See the picture),
internal Wi-Fi,
IR remote & bluster,
internal Capture Card (TV-tuner) nor
Bluetooth in this model.

Note: To tell to truth its bigger brother (L3600) has almost every features (except the PCI slot and Wi-Fi) to be a great HTPC. I saw a used L3600 on the TradeMe just 700NZD with faster CPU, bigger HDD and so on. I would have bought it if I haven't already bought this one.

Motherboard

The Aspire L320 using the ViiV technology with 82801GDH Northbridge and Intel 946GZ (ICH7-DH) Southbridge. There are 2 Audio codecs are coexist in the ICH7-DH chip, but it seems that the AC'97 have been disabled.

The old fashion AC'97 on the 00:1e:2* (Audio) and 00:1e:3 (Modem)
and the newest Intel HD Audio on the 00:1b:00.

Note:00:1e:2 means PCI bus 0, device 27 and function 0.

The mobo (MOtherBOard) of the L320 is a closed product, no any information on the Internet.
So, I had to chase and dig the board a bit to get some information about it.

So, I finally managed to figure out almost all of the secrets of it (See the L320 Block Diagram):

CPU: 1.6GHz Core Duo (LGA 775, E2140, Conroe, 1MB cache, 65nm, 65W) and can be upgreaded up to 2.66GHz Core 2 Duo (E6700, Conroe, 65nm, 65W)
Memory: 1GB -> 2x512MB DIMMs (Dual Channel)
Motherboard (mobo): FI946GZG, 04MJ2302789, Socket 775, mfg: 2007/06/09

o 1xMini-PCI expansion slot
o 1xMini-PCIe expansion slot

NorthBridge: 82946GZ GMCH (Graphic & Memory Controller Hub, 3.3W)

o Integrated Intel 3000 GMA graphics (not for game but MythTV).
o Supporting Dual channel memory

SouthBridge: Intel® 82801GDH ICH7 Digital Home (ICH7-DH) Stepping: A1 (SL8UK)

o AC'97 (disabled in this configuration)
o Intel HD Audio /w Realtek ALC888 HDA Codec /w 7.1+2 Channels
o Vidalia

BIOS bus (LPC or SPI): LPC bus
BIOS type: Phoenix 6.0G
Bios ver: R01-A4 upgraded to R02-B0L (L=Linux)
Bios Chip: Pm49FL004T-3CCJE
Super I/O chip: IT8718F (Not used as there is no any PS/2 Keyborad and Mouse ports)

The Southbridge (ICH7) is divided into 6 logical devices:
ilap@htpc:~$ lspci -tvnn
-[0000:00]-
# Northbridge
+-00.0 Intel Corporation 82946GZ/PL/GL Memory Controller Hub [8086:2970] (Note
+-02.0 Intel Corporation 82946GZ/GL Integrated Graphics Controller [8086:2972]
# Southbridge
#Intel HDA
+-1b.0 Intel Corporation 82801G (ICH7 Family) High Definition Audio Controller [8086:27d8]
# PCI-E root ports (3 of 6)
+-1c.0-[0000:01]--
+-1c.1-[0000:02]--00.0 Atheros Communications Inc. AR242x 802.11abg Wireless PCI-E ... [168c:001c]
+-1c.2-[0000:03]--00.0 Intel Corporation 82573L Gigabit Ethernet Controller [8086:109a]
# USB 4xUHCI and 1xEHCI Controllers
+-1d.0 Intel Corporation 82801G (ICH7 Family) USB UHCI Controller #1 [8086:27c8]
+-1d.1 Intel Corporation 82801G (ICH7 Family) USB UHCI Controller #2 [8086:27c9]
+-1d.2 Intel Corporation 82801G (ICH7 Family) USB UHCI Controller #3 [8086:27ca]
+-1d.3 Intel Corporation 82801G (ICH7 Family) USB UHCI Controller #4 [8086:27cb]
+-1d.7 Intel Corporation 82801G (ICH7 Family) USB2 EHCI Controller [8086:27cc]
# PCI-to-PCI Bridge (PCI 0 to PCI 4)
+-1e.0-[0000:04]--+-06.0 Texas Instruments TSB43AB23 IEEE-1394a... Controller (PHY/Link) [104c:8024]
| \-07.0 Philips Semiconductors SAA7135 Video Broadcast Decoder [1131:7133]
# PCI-to-LPC Bridge, IDE, SATA, SMBus controller
+-1f.0 Intel Corporation 82801GH (ICH7DH) LPC Interface Bridge [8086:27b0]
+-1f.1 Intel Corporation 82801G (ICH7 Family) IDE Controller [8086:27df]
+-1f.2 Intel Corporation 82801GR/GH (ICH7 Family) SATA AHCI Controller [8086:27c1]
\-1f.3 Intel Corporation 82801G (ICH7 Family) SMBus Controller [8086:27da]

See details below:

Sound (Intel HD Audio /w Realtek ALC888 HDA Codec /w 5.1+2 Channels)

It has 6-jack in back and 2-jack in front, so I needed to set the parameter of model of HDA driver to "6stack" or "acer" to get the sound.
See, /usr/share/doc/alsa-base/driver/ALSA-Configuration.txt.gz for details
root@htpc# (S="options snd-hda-intel model=6stack"; T=/etc/modprobe.d/alsa-base; grep "$S" $T > /dev/null || echo "$S" >> "$T")

References

www.acer.com.tw
ark.intel.com

To buy the parts...

I had to enable my eBay account and try to find what components I need to build a Linux based HTPC. I had two choises to fill up the two empty internal slots as I needed one Mini PCI(e) Wi-fi and one Mini PCI(e) capture card to get the L320 to be an HTPC:

Mini PCI Capture card and Mini PCIe Wi-Fi or
Mini PCIe Capture card and Mini PCi Wi-Fi.

Also, I needed an IR remote (to control the HTPC) and Blaster (control
the SKY box), a Wi-FI internal antenna, cable (IPEX/U.FL <-> RP.SMA) for the Wi-Fi card and a u.Fl to BNC pigtail for the 75 ohm TV Antenna Input of the Capture Card.

The prices I found on the eBay were the following:

Mini-PCI Capture Card (/w u.fl to BNC cable) 8.99 USD
Mini-PCIe Wi-Fi 13.99 USD

Mini-PCIe Capture Card (/w areial antenna) 64.99 USD
Mini-PCI Wi-Fi: 10.99 USD

75.98 USD versus 22.98 USD. Which one do you think I choosed?:)
Yep, the winner is the following couple:

AVerMedia M103 (Hybrid DVB-T and NTSC/PAL/SECAM/FM) found as "Mini PCI TV DVB-T PAL Card MCE Asus Acer ie3" on the eBay and it's came with the (cable too)
Atheros AR2425 AR5007EG Mini PCI-E Wi-Fi Card 802.11g.

Finally I have managed to buy the folowing parts:

Hauppauge Media Centre Remote Control Retail Box, Includes IR receiver and IR Blaster (MCE certified) 35.89NZD + 5NZD shipping
Atheros AR2425 AR5007EG MiNi PCI-E WiFi Card 802.11G 13.99USD + free shipping
AverMedia M103 Mini-PCI Hibrid (Analog TV/DVB-T) Card a.k.a Acer TV (MCE certified) /w u.FL to BNC pigtail. 12.49 EUR ~ 15.92USD ~ 31.85 NZD
Tiny USB 2.0 Bluetooth 2.0 EDR Dongle 3.27 USD /w free shipping. 6,35 NZD
wifi U.FL male RP-SMA Pigtail for Mini PCIe Card 20cm. 4.99 USD + Free shipping
10 DBI 2.4GHz WIRELESS WIFI ANTENNA BOOSTER WLAN RP-SMA. 1USD + 3.01USD Shipping

So the cost of the prospective HTPC is ~490 NZD -> ~ 250 USD which I think is a reasonable price for an HTPC. I just needed the L320 to be built as an HTPC.