Jump to content

qeldroma

Members
  • Posts

    422
  • Joined

  • Last visited

Posts posted by qeldroma

  1. As far as i can see, this topic isn't solved till now. So i think, it is time to start a thread to solve this.

    Everyone is invited to give hints :help:

     

    Facts i am pretty sure off:

    • 1. ATI-drivers don't work accelerated on LM9.2 because the dri-library's got unresolved symbols
    • 2. gatos-driver. I am not really sure, but i read, it is only for accelerated 3D, not 2D
       
    • 3. actual XFree-CVS-snapshot. This has accelerated 2D, but is not stable and very hard to compile on Mandrake, don't know why. Perhaps Mandrake's got too many changes in standard-libraries? I don't know :juggle:

    What i want from you is an answer to following questions:

    • 1. Is there an easy and short way, to solve the unresolved symnbols from the driver?
    • 2. An working step-by-step howto to compile a specific CVS-Snapshot on Mandrake or, even better, a ready2use RPM?
    • 3. A way i didn't find out?

    The ideal result of this topic should be a working howto, i will rewrite then, for all modern, NOT WORKING Radeon-cards

  2. It's definitely NOT the monitor-cable or DVI or something else peripherical. The driver in LM9.2 just isn't capable to use this hardware.

     

    Where i know? I got the same card furious3

     

    The only way maybe are this gatos-thing above.

    The ATI-drivers are NOT sources, they fit only on a special version of XFree and a special OS.

    Didn't spend any further time in them, though.

     

    If you've got a faster solution, tell me B)

  3. Ok, now i am on:

    • du: "DiskUsage". Say you want to know how large some directory-structures are... this is it.
    • eject/eject -t: To eject/load a cdrom. Defaults to the first one in alphabetical order. Else you add the /dev/hdX.
    • ntop: same as top, but for network-traffic
    • iptraf: The BEST tool i've found to monitor Network with menues in ncurse to select the wanted options
    • grep: to filter output. A simple "cat /etc/services|grep -i http" will show you all ports concerning http.
    • "|" : this tunnels output to another console-tool, like seen above.
    • ps -ax: shows all processes with applicationname and pid, etc...
    • /proc: You want to know everything about your smart machine like "sysinfo"? Here you are. For every information is a file which is readable (e.g. cpuinfo, filesystems, meminfo, modules, etc...)
    • wget: wanna mirror a http/ftp-directory? "wget -r http://some-url.org/dummy/" will fetch all files in the path "/dummy" on "some-url.org" on your harddisk recursively.
    • /dev/zero: wanna some zeros? Just copy this file to the file you want to fill with zeros. Be aware, it never stopps ;-) Nice way to delete a partition definetely. Just count the zeros with a for-loop until you filled all available bytes.
    • /dev/null: this is a file where nothing is happening, it just eats everything. E.g.: "make bzImage >/dev/null" will copy every compiling-output of make to /dev/null, what means the screen keeps clean.
    • route: shows all network-concerned routes/gateways

    For what do you need this all? Well, a small example:

     

    #!/bin/bash
    echo "Progress:"
    CPU=`cat /proc/cpuinfo|grep "model name"|cut -d":" -f2`;
    MHz=`cat /proc/cpuinfo|grep "MHz"|cut -d":" -f2`;
    cmdLine=`cat /proc/cmdline`;
    modules=`cat /proc/modules|cut -d " " -f1`;
    system=`cat /proc/version`;
    echo -n "#"
    echo "##############################">/var/log/sysinfo
    echo "#### System Informations #####">>/var/log/sysinfo
    echo "##############################">>/var/log/sysinfo
    echo >>/var/log/sysinfo
    echo -n "#"
    
    echo "1. You're system is a $CPU with $MHz MHz.">>/var/log/sysinfo
    echo >>/var/log/sysinfo
    echo -n "#"
    
    echo "2. You're commandline for boot was '$cmdLine'.">>/var/log/sysinfo
    echo >>/var/log/sysinfo
    echo -n "#"
    
    echo "3. You're OS is '$system'.">>/var/log/sysinfo
    echo >>/var/log/sysinfo
    echo >>/var/log/sysinfo
    echo -n "#"
    
    echo "Following all loaded modules with their options:">>/var/log/sysinfo
    echo -n "#"
    for module in $modules; do
       echo "Found Module '$module':">>/var/log/sysinfo
       modinfo $module >>/var/log/sysinfo
       echo >>/var/log/sysinfo
       echo -n "#"
    done
    echo
    echo "finished!"
    read -p "For results press a key."
    clear
    cat /var/log/sysinfo

     

    What you'll see is a progress-meter, after that the result. In /var/log/sysinfo is a file with the same contents.

     

    To be continued.....

     

    :headbang:

  4. Some Months ago, i tested many filesystems under LM9.1 and figured out the performance and CPU-impact. Look at My Webpage (No panic, it's english ;-)).

     

    Another thing is information in general. First of all, you have to seperate simple from professional filesystems. Under "simple" you will find:

    • FAT: This is the primary Windows3.xx/95/98/ME/XP-Home filesystem, under NT/2k/XP there is additionally NTFS. It just seperates the space in blocks and every block is addressed. This makes it fast, but unreliable. There are no further security or backup-functionalities. It is very slow on huge partitions, what lead to users, thinking that they HAVE TO part their disk in thousands of partitions for performance reasons :wall:
      This filesystem is the reason for defragmentation-tools, because it puts the data on the disk without any logic behind. Just takes the biggest free block and that's it.
    • ext2: This is the basic linux FS, similar to FAT but with unix-security-stamp. Because it is born under unix/linux, it is better using space on disk and of course, it is optimized for bigger partitions.

    On the other hand are professional FS. One VERY important feature is journalling, what means, that every action is packed in a transaction and this transactions are logged. So after an error, the system can restore the data according to the log. Another thing is handling of huge storage. Not every FS is capable to do this performant. Think of the CPU-usage. A complex FS needs a lot of management in the behind, so a small CPU will slow down the fs:

    • NTFS: This Windows-FS has additional security-features, but is NOT journalling (or?). It takes about 10-20% of harddisk for a MFT-Block in which the disk is managed. It is much faster than fat, linux can only safely read it, writing support is pre-alpha.
    • ext3: This is nothing really new. It is just a ext2 with an additional information-handling-routine for journalling. So you can use it even without ext3 support in kernel. After an error, it is very slow (20Gig, 7 minutes) in recovering. Because it is based on ext2, many users are using it.
    • ReiserFS: This is an extraordinary FS. All Data is stored in a binary tree-modell, NOT concatenated in blocks. This makes it VERY performant on small files, what most files under linux are. For understanding, think of a pyramid of champagne glasses, one stands on two others, and you fill them from the top. That's simplified the way it works. If you understand that, you will notice, that it can't fragment, because it is based on fragmentation :lol2:. It is relative fast in repairing, but no speed-wonder. All-in-All the perfect FS for small-filed partitions, like /etc, /home, /var, etc...
    • XFS: This is new since about one year. It is a UNIX-based FS which is VERY professional. It has a whole bundle of extra features that are interesting for admins of storage-solutions. Rumours tell about it beeing very fast with huge files, i can't tell this, because it is as fast as ReiserFS as far as i can see. VERY good is repairing-time: it takes only seconds for my 20Gig-partition after an failure for restoring it.
    • JFS: As professional as XFS, but i only tried it once and had problems (a year ago). Perhaps it is stable by now. Rumours tell that it is not stable by now, don't know further.
    • Others: Many others are just for compatibility-reasons with unix-os.

    Be aware that with every OS you are using, you are enlarging the kernel and reducing resources. That's why i only use three FS (FAT for windows, ext2 for boot and XFS for the rest).

     

    Critics and additions are welcome :cheesy:

  5. It's a possibility only available in BASH and Kornshell.

     

    If you've got a variable like $var, you can echo the filtered variable:

     

    echo ${var/...}

     

    Where ... is the content to be removed. There are lot's of additional possibilities, like remove all until end/begin of string depending on a sequence.

    Ex.:

    var = "Hallo du Freak!"
    
    var=`echo ${var/ Freak}`
    
    echo $var

     

    Leads to: "Hallo du!"

     

    One disadvantage is that this way is by far NOT so complex like sed or awk. Only simple things can be done like this...

  6. Ok, did it.

    Now my machine boots the default kernel, but mounts the root via initrd from my server. So my little ITX-Multimedia-PC is without any disks :-)

     

    The easiest way is to take a fat partitition and syslinux' it. Then install the kernel and all necessary modules in the initrd. Voilà

     

    (Not so easy to find in the initrd-things, but that's another story)

  7. I want to build a kernel for my epia-Motherboard and need some not-standard kernel-features, can't use the default one.

    So i need to compile one. Because VIA only delivers prebuild binaries (modules) for their hardware, i am forced to build a kernel that can use them.

     

    Facts:

    1. Theese modules are build for LM9.1-kernel (2.4.21)

    2. I need to turn on "Root on nfs" because diskless client

    3. I need them for X11 !!

    4. It will get a multimediastation, if i can get it working

     

    Option:

    1. Can someone give me a RPM for the BETA XFree86-4.3.99?? It has via-support...

    2. I can't get it compiled, many things missing

  8. I use several remote RPM-sites as sources in URPMI.

    Because they are changing often, i want to see in a type of list, which of my installed RPMs needs/can be updated.

     

    Ex: Installed is SUPERKARAMBA v0.17

     

    If i am trying "urpmi superkaramba" and the actual version is 0.24, urpmi doesn't recognize this. I have manually to "urpme" it, before i can "urpmi"ze it.

     

    Perfect would be a script or short-step-way to do a "update world".

     

    Any ideas?

×
×
  • Create New...