Re: [Xen-devel] igb pciback and e1000e pciback not work.

[Pasi Kärkkäinen <pasik@xxxxxx>]

On Mon, Sep 06, 2010 at 12:58:48PM +0800, ?????? wrote:
>    Thank you for your advise.
>    I update the kernel and xen to the newest version. Now both Dom0 and
>    DomU's kernels are 2.6.32.21-xen

Ok.

>    still the e1000e driver in DomU complains about the "Detected Tx Unit
>    Hang", but there is no kernel oops.Ã
>    It seems the new kernel has fixed a bug about pciback.
>    Maybe I should change theà /etc/xen/xend-pci-permissive.sxp and
>    /etc/xen/xend-pci-quirks.sxp file ,because I found this in Dom0 dmesg:
>    [ Ã 385.007473] pciback 0000:09:00.1: Driver tried to write to a read-only
>    configuration space field at offset 0xe8, size 2. This may be harmless,
>    but if you have problems with your device:
>    [ Ã 385.007475] 1) see permissive attribute in sysfs
>    [ Ã 385.007475] 2) report problems to the xen-devel mailing list along
>    with details of your device obtained from lspci.
>    note: 09:00.0 is the pci interface I move to DomU .
>    But I don't know how to change this.
>    I remove the e1000e driver in Dom0 , still it can't fix the problem.
>

You shouldn't use the device (at all) in dom0 when you're
passing it thru to domU!

Some questions:

- Is that a multiport nic?
- Are you using another nic port in dom0?
- Do you have the "iommu=soft swiotlb=force" options for the domU kernel? 
- How much memory does your domU have? 


-- Pasi


>    Anyone knows about that?
>    thanks.
>    2010/9/5 Pasi KÃârkkÃâinen <[1]pasik@xxxxxx>
> 
>      On Sun, Sep 05, 2010 at 10:55:35AM +0800, ?????? wrote:
>      > Ã  Ã I'm using Xen 4.01 and jeremy's pvops kernel, the kernel version
>      is
>      > Ã  Ã 2.6.32.18.
>      > Ã  Ã At first I want to test the performance of pci passthrough Ã*
>      when it comes
>      > Ã  Ã to the network card 82571EB
>      >
>      > Ã  Ã I'm using the Jeremy's kernel 2.6.32.18 as DomU kernel too.
>      >
>      > Ã  Ã And I found if the interface is not connected, I mean the network
>      card
>      > Ã  Ã receives no packet, ( the network card doesn't connect to the
>      switch ),
>      > Ã  Ã it's fine. when I connect it to a switch , I got a lot of
>      "Detected Tx
>      > Ã  Ã Unit Hang" in dmesg of the DomU kernel
>      > Ã  Ã and a kernel oops saying:
>      >
>      > Ã  Ã Pid: 0, comm: swapper Not tained 2.6.32.18-xen #4
>      > Ã  Ã warn_slowpath_fmt+0x41/0x50
>      >
>      > Ã  Ã the ooops image is in the attachment of this post.
>      >
>      > Ã  Ã I've checked the mailing list, some one said it's about the
>      e1000e driver
>      > Ã  Ã with a wrong
>      > Ã  Ã config of the EEPROM.
>      >
>      > Ã  Ã So I switch to use the network card 82575EB with igb driver
>      > Ã  Ã It seems to work well,Ã*
>      > Ã  Ã until I find it can't receive any packets in DomU.
>      > Ã  Ã And I see a lot of packets dropped when I ifconfig these
>      interfaces.
>      > Ã  Ã Anyone knows how to deal with this ?
>      >
> 
>      Hello,
> 
>      The latest kernel version in xen/stable-2.6.32.x branch is 2.6.32.21.
>      Dunno if it has any fixes that are relevant to you.
> 
>      Anyway, did you read:
>      [2]http://wiki.xensource.com/xenwiki/XenPCIpassthrough
> 
>      -- Pasi
> 
>      _______________________________________________
>      Xen-devel mailing list
>      [3]Xen-devel@xxxxxxxxxxxxxxxxxxx
>      [4]http://lists.xensource.com/xen-devel
> 
>    --
>    hepeng
>    ICT
> 
> References
> 
>    Visible links
>    1. mailto:pasik@xxxxxx
>    2. http://wiki.xensource.com/xenwiki/XenPCIpassthrough
>    3. mailto:Xen-devel@xxxxxxxxxxxxxxxxxxx
>    4. http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel