Difference between revisions of "Desktop Setup - Ubuntu 10.04"

From Vlsiwiki
Jump to: navigation, search
(Installing CUDA 3.2 (incomplete))
Line 176: Line 176:
 
** CUDA Toolkit for Ubuntu Linux 10.04 http://developer.download.nvidia.com/compute/cuda/3_2/toolkit/cudatoolkit_3.2.9_linux_64_ubuntu10.04.run
 
** CUDA Toolkit for Ubuntu Linux 10.04 http://developer.download.nvidia.com/compute/cuda/3_2/toolkit/cudatoolkit_3.2.9_linux_64_ubuntu10.04.run
 
Move these files to your home directory.
 
Move these files to your home directory.
 +
  
 
* Blacklist some kernel modules
 
* Blacklist some kernel modules
Line 185: Line 186:
 
  blacklist nvidiafb
 
  blacklist nvidiafb
 
  blacklist rivatv
 
  blacklist rivatv
 +
 +
 
*Get rid of any installed NVIDIA drivers:
 
*Get rid of any installed NVIDIA drivers:
 
  sudo apt-get --purge remove nvidia-*  
 
  sudo apt-get --purge remove nvidia-*  
 +
  
 
*Reboot the machine. X11 will not start. Switch to a terminal (CTRL + ALT + F1), and login.
 
*Reboot the machine. X11 will not start. Switch to a terminal (CTRL + ALT + F1), and login.
 +
 +
 
* Stop X11
 
* Stop X11
 
  sudo service gdm stop
 
  sudo service gdm stop
 +
  
 
* Install the downloaded device driver first.  
 
* Install the downloaded device driver first.  
 
  sudo devdriver_3.2_linux_64_260.24.run
 
  sudo devdriver_3.2_linux_64_260.24.run
**Accept the license agreement.
+
*Accept the license agreement.  
**Install NVIDIA's 32-bit compatibility OpenGL libraries? Answer: YES
+
*Install NVIDIA's 32-bit compatibility OpenGL libraries? Answer: YES
**Would you like to run the nvidia-xconfig utility to automatically update your X configuration file so that the NVIDIA X driver will be used when you restart X? Answer: YES  
+
*Would you like to run the nvidia-xconfig utility to automatically update your X configuration file so that the NVIDIA X driver will be used when you restart X? Answer: YES  
  
* Install CUDA
+
 
 +
* Install the CUDA toolkit (nvcc etc. )
 
  sudo cudatoolkit_3.2.9_linux_64_ubuntu10.04.run  
 
  sudo cudatoolkit_3.2.9_linux_64_ubuntu10.04.run  
 
Choose default options. CUDA will be installed in the /usr/local/cuda directory. When the installation is complete, you will be asked to make changes to your INCLUDE and LD_LIBRARY_PATH env. variables.
 
Choose default options. CUDA will be installed in the /usr/local/cuda directory. When the installation is complete, you will be asked to make changes to your INCLUDE and LD_LIBRARY_PATH env. variables.
 +
  
 
* Reboot and login.  
 
* Reboot and login.  
 +
  
 
* OPTIONAL: Download GPU Computing SDK code samples http://developer.download.nvidia.com/compute/cuda/3_2/sdk/gpucomputingsdk_3.2_linux.run. This will be installed in your home directory by default.
 
* OPTIONAL: Download GPU Computing SDK code samples http://developer.download.nvidia.com/compute/cuda/3_2/sdk/gpucomputingsdk_3.2_linux.run. This will be installed in your home directory by default.

Revision as of 20:34, 13 October 2010

Currently still testing the distribution, so please report any problems to Ian and Jose!!

  • 0 - Install Ubuntu 10.04 x86_64

NOTE: ONLY if you have trouble with Step 4 of the installation process, the partition selection, open a terminal in the trial mode and run:

sudo dmraid -r -E /dev/sda

NOTE: When prompted for a username, choose one that is NOT your SOE login! For example "ian_local" rather than my SOE login of "ianlee1521". Otherwise you will not be able to login to your computer after you setup LDAP.

NOTE: When prompted for a machine name, enter the name that machine is supposed to be (mascd1, omelette, quiche, etc)

Installation should complete as normal.




  • 1 - Enable root account

Login to the GUI with your local login, then open a terminal (Applications -> Accessories -> Terminal):

sudo su - 
passwd

Enter your new root password for the machine.

Add your local and SOE logins to the sudoers list:

vim /etc/sudoers
  • 2 - Setup the new repo
apt-get install -y emdebian-archive-keyring
echo -e "\ndeb http://www.emdebian.org/debian/ lenny main" | tee -a /etc/apt/sources.list
apt-get update -y
apt-get dist-upgrade -y
  • 3 - Setup the cross package
apt-get install -y apt-cross dpkg-cross libncurses5-dev
  • 4 - Instal ARM packages
apt-get install -y libc6-armel-cross libc6-dev-armel-cross
apt-get install -y binutils-arm-linux-gnueabi gcc-4.3-arm-linux-gnueabi g++-4.3-arm-linux-gnueabi 
apt-get install -y linux-libc-dev-armel-cross
apt-get install -y uboot-mkimage
  • 5 - Install SPARC packages
apt-get install -y libc6-sparc-cross libc6-dev-sparc-cross
apt-get install -y binutils-sparc-linux-gnu gcc-4.3-sparc-linux-gnu g++-4.3-sparc-linux-gnu
apt-get install -y linux-libc-dev-sparc-cross
  • 6 - Adjust sysctl options

mmap problem for qemu

vim /etc/sysctl.d/10-zeropage.conf

Edit the following value:

vm.mmap_min_addr = 4096

increase (add) the inotify watches in /etc/sysctl.conf

echo -e "\nfs.inotify.max_user_watches = 65535" | tee -a /etc/sysctl.conf
  • 7 - Replicate the installation on Master machine (mascd12, dhcp-63-154.cse.ucsc.edu)
ssh <username>@mascd12 dpkg --get-selections > mypackages.txt
dpkg --set-selections < mypackages.txt

"apt-get dselect-upgrade -y" Will give many options and take a long time for the upgrade (about an hour for me).

Select all default options EXCEPT! for ion3: acknowledge the lack of support (non-default option)

apt-get dselect-upgrade -y
  • 8 - Install autofs
mkdir /mada

Get autofs files from another machine

scp <username>@mascd12:/etc/auto.* /etc/
ln -s /mada/users /soe
initctl reload autofs
  • 9 - awk for cadence
apt-get install -y ksh csh original-awk
ln -s /usr/bin/awk /bin/awk
  • 10 - Set the firewall
ufw enable
ufw default deny
ufw reject auth
ufw limit ssh/tcp
  • 11 - Get tempcap from another ubuntu machine (mascd1, mascd12, ...)
scp <username>@mascd12:/usr/lib64/libtermcap.so.2.0.8 /usr/lib64/
ln -s /usr/lib64/libtermcap.so.2.0.8 /usr/lib64/libtermcap.so.2
  • 12 - LDAP

Install

apt-get install -y ldap-auth-client libpam-krb5 krb5-user libsasl2-modules-gssapi-mit

Leave the kerberos config entry blank. You will copy it in the next step.

Create /ect/krb5.conf

scp <username>@mascd12:/etc/krb5.conf /etc/

Copy the ldap certificate

scp <username>@mascd12:/etc/ssl/certs/* /etc/ssl/certs/

Create "/etc/auth-client-config/profile.d/krb-ldap-config"

scp <username>@mascd12:/etc/auth-client-config/profile.d/krb-ldap-config /etc/auth-client-config/profile.d/krb-ldap-config
auth-client-config -a -p krb_ldap

Update ldap.conf

scp <username>@mascd12:/etc/ldap.conf /etc/ldap.conf
scp <username>@mascd12:/etc/ldap/ldap.conf /etc/ldap/ldap.conf

Now you should be able to execute:

ldapsearch -x LLL
  • 13 - Lock SOE people out of your machine :)
scp <username>@mascd12:/etc/security/access.conf /etc/security/access.conf

You will need to add your local login to the machine, or you will lock yourself out:

vim /etc/security/access.conf

Go to very bottom of file and add your local user account to the list below root Example of the last lines of my file (Replace "ian_local" with your local login):

+ : root : ALL
+ : ian_local : ALL
+ : (masc) : ALL
+ : (vlsi) : ALL
-:ALL :ALL 

Now, you should see people with "getent passwd" but, calling "su - cormac" should say "permission denied", but, calling "su - renau" should work.

  • 14 - Restart your computer:
shutdown -r now
  • 15 - Login to the GUI with your SOE login information. If you get a prompt asking for an "LDAP Password" this means that you typed your password incorrectly.
  • 16 - Install NVIDIA Drivers (Please read all instructions before proceeding!)

Open a terminal and type:

sudo su -
wget http://us.download.nvidia.com/XFree86/Linux-x86_64/256.53/NVIDIA-Linux-x86_64-256.53.run
chmod 744 NVIDIA-Linux-x86_64-256.53.run
service gdm stop

Login as root, if not already, and rerun the installation

sudo su - 
./NVIDIA-Linux-x86_64-256.53.run

NOTE: You may get several error messages during the installation including "PreInstall Script has failed", Nouveau is causing problems, etc. These are normal, and following all of the instructions will resolve them.

The installation will fail due to the Nouveau drivers, but it will create a script to disable them. Just restart the computer and reclose X, then go through installation again and it should work as expected.

shutdown -r now

Switch to tty1 (CTRL + ALT + F1), and login as root.

service gdm stop
./NVIDIA-Linux-x86_64-256.53.run

NOTE: NVIIDIA Drivers should have installed properly this time. Make sure that you choose to overwrite the xconfig file.

Restart for the changes to take effect properly:

shutdown -r now
  • 17 - Remove old desktop configuration files (KDE / GNOME / browser)

NOTE: Removing ".config" will delete any saved passwords in your browser, but will fix other problems.

Switch to tty1 (CTRL + ALT + F1), and login with your SOE account.

rm -rf .config .kde* .gconf* .gnome*
  • 18 - Some Minor Patches for things that don't work properly

Disable apparmor

sudo invoke-rc.d apparmor stop
sudo update-rc.d -f apparmor remove

Enable 32 bit compilation

ln -s /usr/lib32/libstdc++.so.6 /usr/lib32/libstdc++.so

Installing CUDA 3.2 (incomplete)

You will need sudo access to install CUDA 3.2. It can only be installed on machines with NVIDIA GPUS.

Move these files to your home directory.


  • Blacklist some kernel modules
sudo vim /etc/modprobe.d/blacklist.conf

Add the following lines to the file

blacklist vga16fb
blacklist nouveau
blacklist rivafb
blacklist nvidiafb
blacklist rivatv


  • Get rid of any installed NVIDIA drivers:
sudo apt-get --purge remove nvidia-* 


  • Reboot the machine. X11 will not start. Switch to a terminal (CTRL + ALT + F1), and login.


  • Stop X11
sudo service gdm stop


  • Install the downloaded device driver first.
sudo devdriver_3.2_linux_64_260.24.run
  • Accept the license agreement.
  • Install NVIDIA's 32-bit compatibility OpenGL libraries? Answer: YES
  • Would you like to run the nvidia-xconfig utility to automatically update your X configuration file so that the NVIDIA X driver will be used when you restart X? Answer: YES


  • Install the CUDA toolkit (nvcc etc. )
sudo cudatoolkit_3.2.9_linux_64_ubuntu10.04.run 

Choose default options. CUDA will be installed in the /usr/local/cuda directory. When the installation is complete, you will be asked to make changes to your INCLUDE and LD_LIBRARY_PATH env. variables.


  • Reboot and login.