Find the answer to your Linux question:
Results 1 to 4 of 4
Hi everybody, I'm a new linux user. I would like to increase the size of the mtu on my gigabit ethernet nic (eth6) for jumbo frames. When I do this ...
  1. #1
    Just Joined!
    Join Date
    Jul 2006
    Posts
    13

    problem with memory and jumbo frames



    Hi everybody,
    I'm a new linux user. I would like to increase the size of the mtu on my gigabit ethernet nic (eth6) for jumbo frames. When I do this (using "mtu" in /etc/network/interfaces) I get stuff like this on the console:

    swapper: page allocation failure. order:2, mode:0x20
    [<c0143449>] __alloc_pages+0x2f9/0x370
    [<c01434e5>] __get_free_pages+0x25/0x40
    [<c0146fe3>] kmem_getpages+0x23/0xd0
    [<c0254b46>] ip_forward+0x1d6/0x2c0
    [<c0147d96>] cache_grow+0xe6/0x1e0
    [<c0148023>] cache_alloc_refill+0x193/0x250
    [<c0148558>] __kmalloc+0x88/0xa0
    [<c022fdc7>] alloc_skb+0x47/0xf0
    [<d0b906c5>] e1000_alloc_rx_buffers+0x65/0x110 [e1000]
    [<d0b902cb>] e1000_clean_rx_irq+0xfb/0x490 [e1000]
    [<c011c05b>] rebalance_tick+0xab/0xc0
    [<d0b8ff2a>] e1000_intr+0x3a/0x90 [e1000]
    [<c01086c9>] handle_IRQ_event+0x49/0x80
    [<c0108a86>] do_IRQ+0xb6/0x1a0
    [<c0106bd0>] common_interrupt+0x18/0x20
    [<c0104030>] default_idle+0x0/0x40
    [<c010405c>] default_idle+0x2c/0x40
    [<c01040eb>] cpu_idle+0x3b/0x50
    [<c01212ce>] printk+0x16e/0x1c0

    right after when I do ifup and it repeats forever every 10 seconds.
    my understanding is that swapper must have something to do with the memory? So what is this, does it matter and do I need to fix anything?
    I'm running debian sarge 3.1r2 on x86 with the 2.6.8-3-686-smp linux kernel.
    Thanks for your help

  2. #2
    Linux User
    Join Date
    Aug 2005
    Location
    Italy
    Posts
    401
    It seems a kernel bug, or at least a kernel failure.

    I suggest to upgrade to last kernel, but this doesn't ensure that problem is solved. Maybe the module e1000 can't handle the MTU value specified...

    Maybe sending that backtrace to the e1000 module author may help.
    When using Windows, have you ever told "Ehi... do your business?"
    Linux user #396597 (http://counter.li.org)

  3. #3
    Just Joined!
    Join Date
    Jul 2006
    Posts
    13
    hum....I don't think I can send this to anybody since this module appears to come from Intel directly (the nic is from Intel and they give very good linux support). I do not really feel like upgrading the kernel, since the one I have is the latest in -stable. Maybe what I will do is simply leave the mtu to 1500. When the newer kernels will be available for -stable, I will try back jumboframes. I am successfully using jumbo frames with this nic on two other servers running windows 2000 server and freebsd....must be a linux memory problem only.

  4. #4
    Just Joined! letmeknowlinux's Avatar
    Join Date
    Mar 2006
    Posts
    82
    It seems to be like memory allocation problem:

    Here there is a discussion regarding this, some people found a workaround but they are not sure about how other applications might effect.
    http://www.gatago.com/linux/kernel/14714295.html

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •