Re: [Xen-devel] Re: [Xen-users] wiki page on 10G SRIOV

[Pasi Kärkkäinen <pasik@xxxxxx>]

On Mon, Jul 19, 2010 at 02:59:52PM -0700, kishore kumar wrote:
>    I am using Ethernet Server Adapater X520 - DA2 (E10G42BTDA) Card.
>    For the Device Id 0x10FB, I was able to see all the 64 Virtual Functions,
>    when I run lspci command.
>    But when the Device ID started showing 0x10D8, I cannot see any Virtual
>    Functions, when I run lspci command.
>

So the question is what happened? How/when did it change device ID? 
Did you flash/upgrade the firmware?

-- Pasi

>    On Mon, Jul 19, 2010 at 2:35 PM, Rose, Gregory V
>    <[1]gregory.v.rose@xxxxxxxxx> wrote:
> 
>      If you have an 82599EB that is using 10D8h for the Virtual Function
>      Device ID then the adapter is a very old, preproduction version or else
>      the correct EEPROM image has been over-written with an invalid image.Ã
>      The correct Device ID value for an 82599EB virtual function is 10EDh.
> 
>      Ã
> 
>      - Greg
> 
>      Ã
> 
>      From: [2]xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>      [mailto:[3]xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Byrne,
>      John (HP Labs)
>      Sent: Monday, July 19, 2010 2:22 PM
>      To: kishore kumar
>      Cc: [4]xen-devel@xxxxxxxxxxxxxxxxxxx; wangalalei;
>      [5]xen-users@xxxxxxxxxxxxxxxxxxx
>      Subject: RE: [Xen-devel] Re: [Xen-users] wiki page on 10G SRIOV
> 
>      Ã
> 
>      IÃ**m a bit confused with your device IDs. The 10fb device is the PCI id
>      for the physical NIC; the 10d8 device is the SRIOV Virtual Function
>      device created by the driver. Ã With the latest xen-unstable, I have
>      successfully used this NIC to create and assign the VF devices to HVM
>      guests. The ixgbevf driver, which binds to the VF devices, expects
>      device id 0x10ed. I simply assumed that the driver was not quite up to
>      date and tweaked things to work. Ã So, if you are just a bit confused,
>      the following might help.
> 
>      Ã
> 
>      I added the following in my modprobe.d directory to create four VF
>      devices (ixgbe.conf, attached):
> 
>      Ã
> 
>      options ixgbe max_vfs=4
> 
>      Ã
> 
>      So, on my machine, I can see the physical NIC and the 4 VFs created.
> 
>      Ã
> 
>      lspci Ã**n | grep 0a:
> 
>      0a:00.0 0200: 8086:10fb (rev 01)
> 
>      0a:10.0 0200: 8086:10d8 (rev 01)
> 
>      0a:10.2 0200: 8086:10d8 (rev 01)
> 
>      0a:10.4 0200: 8086:10d8 (rev 01)
> 
>      0a:10.6 0200: 8086:10d8 (rev 01)
> 
>      Ã
> 
>      I added the ixgbevf driver to both my dom0 and domUs. While it is a
>      simple tweak to add the device id to the driver, I decided to play with
>      modprobe instead and added the following to my modprobe.d directory in
>      both dom0 and domU (ixgbvevf.conf, attached):
> 
>      Ã
> 
>      install ixgbevf /sbin/modprobe --ignore-install ixgbevf; echo "0x8086
>      0x10d8" >/
> 
>      sys/bus/pci/drivers/ixgbevf/new_id
> 
>      Ã
> 
>      If you are working with HVM guests, there has to be a patch to à qemu as
>      well. (Attached) Ã You also have to add Ã**(pci-passthrough-strict-check
>      no)Ã** to your xend-config.sxp. (I donÃ**t think you need the
>      xend-config change for PV guests.)
> 
>      Ã
> 
>      Then it is simply a matter of adding the VF PCI devices to your domU
>      configurations and configuring the ixgbevf devices in the guests.
> 
>      Ã
> 
>      Both the ixgbe and ixgbevf drivers will generate random MAC addresses
>      for the devices. Whatà I wanted to do was to set the MAC addresses for
>      the VF devices in dom0 and then my guestsÃ** networking configuration
>      was much easier. The physical NIC must be up. It doesnÃ**t have to have
>      an IP address, but if it isnÃ**t up the MAC address set in dom0 will be
>      lost. IÃ**ve attached my dom0 rc.local where I set the MAC addresses and
>      assign the devices to pciback.
> 
>      Ã
> 
>      I canÃ**t swear that this is the best way to do this, but this did work.
> 
>      Ã
> 
>      John Byrne
> 
>      Ã
> 
>      Ã
> 
>      From: [6]xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>      [mailto:[7]xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of kishore
>      kumar
>      Sent: Monday, July 19, 2010 10:16 AM
>      To: Pasi KÃârkkÃâinen; [8]xen-devel@xxxxxxxxxxxxxxxxxxx
>      Cc: wangalalei; [9]xen-users@xxxxxxxxxxxxxxxxxxx
>      Subject: [Xen-devel] Re: [Xen-users] wiki page on 10G SRIOV
> 
>      Ã
> 
>      Hello Everyone,
> 
>      Ã
> 
>      I am facing a strange issue with my 10G 82599 PCIe device.
> 
>      Ã
> 
>      The Device Id has to be 0x10FB for SRIOV to function properly, But all
>      of sudden the Device IDÃ started showing me as 0x10D8, and because of
>      this the SRIOV feature is not working and I cannot assign the Virtual
>      Function toà my guest.
> 
>      Ã
> 
>      When I did lspci command it is showing me as Intel Corporation Unknown
>      device 10d8 (rev 01).Ã  Previously it was showing me Intel Corporation
>      82599EB Gigabit Networking with 10FB, and I was able to assign the
>      Virtual Function to my guest.
> 
>      Ã
> 
>      I googled and found out that the device ID 0x10D8 is a 82599
>      Unprogrammed device.
> 
>      Ã
> 
>      Can any one please let me know what is happening here and how to recover
>      back from this?
> 
>      I am totally blocked for 2 days because of this issue.
> 
>      Ã
> 
>      Thank You.
> 
>      On Fri, Jul 16, 2010 at 1:05 AM, Pasi KÃârkkÃâinen <[10]pasik@xxxxxx>
>      wrote:
> 
>      On Fri, Jul 16, 2010 at 02:17:30AM +0000, wangalalei wrote:
>      > Ã  Ã hi all:
>      >
>      > Ã  Ã  Ã  i read intel 82576 card datasheet,and finding it saying that
>      the 82576
>      > Ã  Ã can set the vf's bandwidth,
>      > Ã  Ã but when i test this card,i can not find any parameter to it
>      > Ã  Ã  Ã  any one else notice this problem? or anyone knows whether the
>      82576
>      > Ã  Ã card can control its vf's bandwidth?
>      >
> 
>      I think bandwidth management is a work-in-progress by Intel..
> 
>      -- Pasi
> 
>      > Ã  Ã alaleiwang
>      >
>      > Ã
>      Ã 
> --------------------------------------------------------------------------
>      >
>      > Ã  Ã Date: Thu, 15 Jul 2010 18:35:57 -0700
>      > Ã  Ã Subject: Re: [Xen-users] wiki page on 10G SRIOV
>      > Ã  Ã From: [11]bodkekumar@xxxxxxxxx
>      > Ã  Ã To: [12]pasik@xxxxxx
>      > Ã  Ã CC: [13]xen-users@xxxxxxxxxxxxxxxxxxx
>      >
>      > Ã  Ã Yeah, I configured 64 Virtual Functions in DOM-0 Ã and passthru 1
>      VF to
>      > Ã  Ã assign it to the guest. Ã It worked for me.
>      >
>      > Ã  Ã I assign static IP to my guest and tried to ping the remote
>      machine. It
>      > Ã  Ã worked for me.
>      >
>      > Ã  Ã But I am wondering that ping to DOM-0 from guest makes any sense
>      here?
>      > Ã  Ã Can you let me know on this? Ã ping worked though from guest to
>      DOM-0 as
>      > Ã  Ã well.
>      >
> 
>      > Ã  Ã On Thu, Jul 15, 2010 at 1:11 PM, Pasi KÃ*Ã*ÂrkkÃ*Ã*Âinen
>      <[1][14]pasik@xxxxxx>
> 
>      > Ã  Ã wrote:
>      >
>      > Ã  Ã  Ã On Thu, Jun 24, 2010 at 10:56:28AM -0700, kishore kumar wrote:
>      > Ã  Ã  Ã > Ã  Ã Hi All,
>      > Ã  Ã  Ã >
>      > Ã  Ã  Ã > Ã  Ã Can any one please let me know if there is a WIKI page
>      which talks
>      > Ã  Ã  Ã about
>      > Ã  Ã  Ã > Ã  Ã how to configure SRIOV in Xen and use Physical
>      Functions à and
>      > Ã  Ã  Ã Virtual
>      > Ã  Ã  Ã > Ã  Ã Functions ?
>      > Ã  Ã  Ã >
>      > Ã  Ã  Ã > Ã  Ã I searched xen wiki, but could not find.
>      > Ã  Ã  Ã >
>      >
>      > Ã  Ã  Ã Each virtual function should be a separate PCI device, so you
>      can
>      > Ã  Ã  Ã passthru
>      > Ã  Ã  Ã them like any other normal PCI device.
>      >
>      > Ã  Ã  Ã Did you get it to work?
>      > Ã  Ã  Ã -- Pasi
>      >
> 
>      > Ã
>      Ã 
> --------------------------------------------------------------------------
>      >
>      > Ã  Ã ÃÅ**ÃÂÃ*ÂÃÂ+ÃÅ**ÃÂÃÅÃÂ+ÃÂ*ÃÂÃÂÃÂÃÂ
>      ÃÅ*ÃÂÃÅà *ÃÅÃ*ÃÂÃÅ*ÃÅÃÂ*ÃÂÃÅÃÅÃÅ,ÃÅ**ÃÅ*ÃÂMSNÃÂÃÅÃÂÃâÃ* ÃÅ**ÃÂÃÂ*!
>      > à  à [2]ÃÂÃÂ*ÃÂ*ÃÂÃâÃÅ*ÃÅÃ*Ã*ÃÂÃ**
> 
>      >
>      > References
>      >
>      > Ã  Ã Visible links
>      > Ã  Ã 1. mailto:[15]pasik@xxxxxx
> 
>      > Ã  Ã 2. [16]http://3g.msn.cn/
> 
>      Ã
> 
> References
> 
>    Visible links
>    1. mailto:gregory.v.rose@xxxxxxxxx
>    2. mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>    3. mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>    4. mailto:xen-devel@xxxxxxxxxxxxxxxxxxx
>    5. mailto:xen-users@xxxxxxxxxxxxxxxxxxx
>    6. mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>    7. mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>    8. mailto:xen-devel@xxxxxxxxxxxxxxxxxxx
>    9. mailto:xen-users@xxxxxxxxxxxxxxxxxxx
>   10. mailto:pasik@xxxxxx
>   11. mailto:bodkekumar@xxxxxxxxx
>   12. mailto:pasik@xxxxxx
>   13. mailto:xen-users@xxxxxxxxxxxxxxxxxxx
>   14. mailto:pasik@xxxxxx
>   15. mailto:pasik@xxxxxx
>   16. http://3g.msn.cn/

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


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