Announcement

Collapse
No announcement yet.

ndiswrapper w/4kb kernel

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • ndiswrapper w/4kb kernel

    anyone tried ndiswrapper with kernel enabled 4kb stack stuff?
    or anyone got any patch for this problem?

    i tried searching in several forums but still no possible solution

    thanks gurus

    Code:
    D0 N0T read the b00k
    D0 N0T read the c0de
    DO code your b00k
    D0 N0T b00k your code

  • #2
    Originally posted by h0n3ym4k
    anyone tried ndiswrapper with kernel enabled 4kb stack stuff?
    or anyone got any patch for this problem?
    I don't know enough about the problem to answer your question.

    Could you provide links to the problem you and others are having?

    Is this an issue with the minimum tcp stack size for the tcp sliding window size as being a default of 4k is too small?

    Have you examined your /proc information to see if you can modify the minimum?

    /proc/sys/net/ipv4/tcp_wmem ?
    tcp_wmem - vector of 3 INTEGERs: min, default, max
    min: Amount of memory reserved for send buffers for TCP socket.
    Each TCP socket has rights to use it due to fact of its birth.
    Default: 4K
    How to examine?
    # cat /proc/sys/net/ipv4/tcp_wmem

    /proc/sys/net/ipv4/tcp_rmem ?
    tcp_rmem - vector of 3 INTEGERs: min, default, max
    min: Minimal size of receive buffer used by TCP sockets.
    It is guaranteed to each TCP socket, even under moderate memory
    pressure.
    Default: 8K
    How to examine?
    # cat /proc/sys/net/ipv4/tcp_rmem

    If your kernel defaults to 4k, have you tried changing the minimum from 4k to 8k or whatever you want?
    # echo "8192 16384 131072" >> /proc/sys/net/ipv4/tcp_wmem
    # echo "8192 87380 174760" >> /proc/sys/net/ipv4/tcp_rmem
    ?

    While you are at it, why not inspect these too:
    /proc/sys/net/ipv4/tcp_mem
    /proc/sys/net/core/wmem_max
    /proc/sys/net/core/rmem_max

    Of course, don't just change them without understanding what you are doing.

    Maybe I just dont understand the problem. More background would help.
    [Added content]
    For more information on kernel /proc stuff, check out these files in your kernel source tree. (Assuming you are in the base of your linux kernel source tree...)
    Documentation/filesystems/proc.txt
    Documentation/networking/proc_net_tcp.txt
    Documentation/networking/ip-sysctl.txt

    I AM THE GREAT CORNHOLIO! Heh-heh. Whoah! EVERYONE COME OUT WITH YOUR PANTS DOWN! Yeah! That is pretty cool!
    Last edited by TheCotMan; March 6, 2005, 00:29.

    Comment


    • #3
      Gotta love TheCotMan...

      Originally posted by TheCotMan
      I don't know enough about the problem to answer your question.
      Followed by 35 lines of answer...
      Originally posted by TheCotMan
      Could you provide links to the problem you and others are having?

      Is this an issue with the minimum tcp stack size for the tcp sliding window size as being a default of 4k is too small?

      Have you examined your /proc information to see if you can modify the minimum?

      /proc/sys/net/ipv4/tcp_wmem ?
      tcp_wmem - vector of 3 INTEGERs: min, default, max
      min: Amount of memory reserved for send buffers for TCP socket.
      Each TCP socket has rights to use it due to fact of its birth.
      Default: 4K
      How to examine?
      # cat /proc/sys/net/ipv4/tcp_wmem

      /proc/sys/net/ipv4/tcp_rmem ?
      tcp_rmem - vector of 3 INTEGERs: min, default, max
      min: Minimal size of receive buffer used by TCP sockets.
      It is guaranteed to each TCP socket, even under moderate memory
      pressure.
      Default: 8K
      How to examine?
      # cat /proc/sys/net/ipv4/tcp_rmem

      If your kernel defaults to 4k, have you tried changing the minimum from 4k to 8k or whatever you want?
      # echo "8192 16384 131072" >> /proc/sys/net/ipv4/tcp_wmem
      # echo "8192 87380 174760" >> /proc/sys/net/ipv4/tcp_rmem
      ?

      While you are at it, why not inspect these too:
      /proc/sys/net/ipv4/tcp_mem
      /proc/sys/net/core/wmem_max
      /proc/sys/net/core/rmem_max

      Of course, don't just change them without understanding what you are doing.

      Maybe I just dont understand the problem. More background would help.
      [Added content]
      For more information on kernel /proc stuff, check out these files in your kernel source tree. (Assuming you are in the base of your linux kernel source tree...)
      Documentation/filesystems/proc.txt
      Documentation/networking/proc_net_tcp.txt
      Documentation/networking/ip-sysctl.txt

      I AM THE GREAT CORNHOLIO! Heh-heh. Whoah! EVERYONE COME OUT WITH YOUR PANTS DOWN! Yeah! That is pretty cool!
      perl -e 'print pack(c5, (41*2), sqrt(7056), (unpack(c,H)-2), oct(115), 10)'

      Comment


      • #4
        Originally posted by Chris
        Gotta love TheCotMan...
        [snip]
        Followed by 35 lines of answer...
        Yeah. That kind of thing happens when you are a Level 1 Ninja stabbing the the dark with a sharpened 2x4... Sure, I can wear black, but walking through walls is beyond my grasp.

        I just took a guess... Sure, I am probably wrong, but it was worth a try. :-)

        [Meanwhile, I just wait for more information from the person posting the problem to better address the question.]
        Last edited by TheCotMan; March 6, 2005, 11:59.

        Comment


        • #5
          Damn I wish I had those 1337 skills, not being able to understand a question but is able to answer the question. I'm not talking about a half ass answer either, cheers to you CotMan.

          Comment


          • #6
            Originally posted by mikedc1760
            Damn I wish I had those 1337 skills, not being able to understand a question but is able to answer the question. I'm not talking about a half ass answer either, cheers to you CotMan.
            Heh. My skills are not leet. There are others on the wireless section who have much better skills than me. There are people here who can walk through walls while I end up walking into them. ;-)

            There are a few things that could be the issue with this original question. One has to do with min window size for tcp connections in Linux, another has to do with allocation of memory by the kernel for processes (chunks and sizes allocated) and I am sure there are others that are also in kernel space that may be related. ("4k kernel stuff"" is ambiguous-- people might be surprised as to how often "4k" is a configurable option for different kernel option in Linux.)

            Instead of just asking more questions, I figured I'd try to address a common issue the same time I asked more questions to better understand the real problem.

            If the first answer is/was correct, then luck leads to efficiency. If the first answer is not correct, then more answers add more value to this thread for others who have the same problem but who require different solutions.

            Meanwhile, I continue to stumble around in the dark with this 2x4 hoping for illumination in the form of more information about the problem.

            Heh-heh, 2x4 are made of *wood*. heh-he heh-heh hrm. ummm. That's pretty cool.

            Comment

            Working...
            X