Yes,  Greg is correct.  SR-IOV _requires_ MSI-X.  If there is no MSI-X you 
cannot get SR-IOV to work correctly.

It is that simple.   Sorry.

Cheers,
John


> -----Original Message-----
> From: mhban [mailto:[email protected]]
> Sent: Monday, June 17, 2013 3:00 AM
> To: Rose, Gregory V; Skidmore, Donald C
> Cc: Ronciak, John; Kirsher, Jeffrey T; e1000-
> [email protected]; Brandeburg, Jesse
> Subject: RE: [E1000-devel] Regarding 82599 SR-IOV support in MSI
> 
> Hi Greg,
> 
> By the way, I have Intel Centerton (S1200) board also. As far as I know
> Centerton support only MSI but X520 IOV seems work well in this board.
> Am I mistaken?
> 
> Regards,
> Minho Ban
> 
> -----Original Message-----
> From: Greg Rose [mailto:[email protected]]
> Sent: Saturday, June 15, 2013 6:23 AM
> To: Skidmore, Donald C
> Cc: mhban; Ronciak, John; Kirsher, Jeffrey T; e1000-
> [email protected]; Brandeburg, Jesse
> Subject: Re: [E1000-devel] Regarding 82599 SR-IOV support in MSI
> 
> On Fri, 14 Jun 2013 17:09:49 +0000
> "Skidmore, Donald C" <[email protected]> wrote:
> 
> > We haven't implemented this as you know so I haven't looked in to it
> > at all in detail but off the top of my head:
> >
> > There is only one interrupt with MSI per say, as John mentioned.  So
> > you will either have to only get all interrupts in the PF and have it
> > tell the VF who they belonged to, nasty latency if doable at all.  Or
> > have all VF's and the PF interrupted for any interrupt that was
> > destine for any VF or the PF.  This second choice would be a lot of
> > overhead and also would mean any VF would know when other VF's were
> > getting interrupts (security issue?).
> 
> The VF cannot send MSI interrupts, period.  It is a HW limitation -
> they only support MSIX.  It is not possible to use MSI with Intel VF
> devices in any way or manner.
> 
> - Greg
> 
> >
> > Thanks,
> > -Don Skidmore <[email protected]>
> >
> > > -----Original Message-----
> > > From: mhban [mailto:[email protected]]
> > > Sent: Thursday, June 13, 2013 7:50 PM
> > > To: Ronciak, John; Skidmore, Donald C; Kirsher, Jeffrey T
> > > Cc: [email protected]; Brandeburg, Jesse
> > > Subject: RE: [E1000-devel] Regarding 82599 SR-IOV support in MSI
> > >
> > > Dear John and Don,
> > > Thanks for kind reply.
> > >
> > > This ARM SoC doesn't have additional features for virtualization
> > > like VT-d. And I'm not going to use SR-IOV for virtualization. I
> > > just want to do some experimental VF test only with limited 2 or 4.
> > >
> > > BTW, I have one more question about this.
> > > Do you mean root-port can receive interrupt from VFs that are
> > > configured as MSI mode?
> > > If so, is it possible to differentiate between them? For example,
> by
> > > reading flag registers again or something like that.
> > >
> > >
> > >
> > > -----Original Message-----
> > > From: Ronciak, John [mailto:[email protected]]
> > > Sent: Friday, June 14, 2013 12:13 AM
> > > To: Skidmore, Donald C; mhban; Kirsher, Jeffrey T
> > > Cc: [email protected]; Brandeburg, Jesse
> > > Subject: RE: [E1000-devel] Regarding 82599 SR-IOV support in MSI
> > >
> > > Hello,
> > >
> > > No it's not possible.  MSI-X is needed so that interrupts can be
> > > routed to each guest.  With MSI there is only a single interrupt so
> > > there is no way to route it to multiple VF's.  MSI-X is a
> > > requirement for SR-IOV support.  So don's comments below are
> > > correct.  Sorry.
> > >
> > > Does this system support VT-d or something like it for the address
> > > translations?  Just curious.
> > >
> > > Cheers,
> > > John
> > >
> > > Cheers,
> > > John
> > >
> > >
> > > > -----Original Message-----
> > > > From: Skidmore, Donald C [mailto:[email protected]]
> > > > Sent: Thursday, June 13, 2013 7:17 AM
> > > > To: mhban; Kirsher, Jeffrey T
> > > > Cc: [email protected]; Brandeburg, Jesse
> > > > Subject: Re: [E1000-devel] Regarding 82599 SR-IOV support in MSI
> > > >
> > > > Hey Minho,
> > > >
> > > > I believe that it would be problematic at best since as MSI uses
> > > > identical address for all vectors in a Function and
> differentiates
> > > > vectors in a Function only by LSBs in data.
> > > >
> > > > Thanks,
> > > > -Don Skidmore <[email protected]>
> > > >
> > > > > -----Original Message-----
> > > > > From: mhban [mailto:[email protected]]
> > > > > Sent: Thursday, June 13, 2013 12:40 AM
> > > > > To: Kirsher, Jeffrey T
> > > > > Cc: [email protected]; Brandeburg, Jesse
> > > > > Subject: [E1000-devel] Regarding 82599 SR-IOV support in MSI
> > > > >
> > > > > Dear maintainer and all SR-IOV experts,
> > > > >
> > > > > I'm developing a board which is based on ARM SoC which support
> > > > > PCI Express root-port.
> > > > > I'm trying to install and activate SR-IOV with Intel X520-SR2
> on
> > > > > this board(/w Linux). Up to now, ixgbe and ixgbevf driver is
> > > > > loaded successfully but driver prints error when I try to do
> > > > > 'ifup'.
> > > > >
> > > > >        "PF still in reset state".
> > > > >
> > > > > While debugging it I found one important article from Intel
> > > > > community site,
> > > > >
> > > > >       http://communities.intel.com/message/158673#158673
> > > > >
> > > > > I realized Intel 82599 require MSI-X for SR-IOV not MSI, but
> the
> > > > > ARM SoC only support MSI.
> > > > > What I want to know is if it is possible to activate SR-IOV
> with
> > > > > MSI (not MSI- X). I know ixgbevf officially doesn't support MSI
> > > > > for SR-
> > > > IOV
> > > > > but I just want to use it for experimental test on the ARM
> > > > > environment. According to the article, it seems there was no
> > > > > hardware limitation which  prevent the use of MSI. If so, can I
> > > > > make work SR-IOV using MSI instead of MSI-X by little re- work
> > > > > ixgbevf(or ixgbe
> > > > also) driver?
> > > > >
> > > > > Please shed some light on this.
> > > > >
> > > > > Best regards,
> > > > > Minho Ban
> > > > >
> > > > > SAMSUNG Electronics, Co., Ltd.
> > > > > (Office) +82-31-301-8489  (Mobile) +82-10-8725-8864
> > > > >
> > > > >
> > > > > ---------------------------------------------------------------
> -
> > > > > ----
> > > > > -
> > > > -
> > > > > -------- This SF.net email is sponsored by Windows:
> > > > >
> > > > > Build for Windows Store.
> > > > >
> > > > > http://p.sf.net/sfu/windows-dev2dev
> > > > > _______________________________________________
> > > > > E1000-devel mailing list
> > > > > [email protected]
> > > > > https://lists.sourceforge.net/lists/listinfo/e1000-devel
> > > > > To learn more about Intel&#174; Ethernet, visit
> > > > > http://communities.intel.com/community/wired
> > > >
> > > > -----------------------------------------------------------------
> -
> > > > ----
> > > > -
> > > > -------
> > > > This SF.net email is sponsored by Windows:
> > > >
> > > > Build for Windows Store.
> > > >
> > > > http://p.sf.net/sfu/windows-dev2dev
> > > > _______________________________________________
> > > > E1000-devel mailing list
> > > > [email protected]
> > > > https://lists.sourceforge.net/lists/listinfo/e1000-devel
> > > > To learn more about Intel&#174; Ethernet, visit
> > > > http://communities.intel.com/community/wired
> >
> >
> > ---------------------------------------------------------------------
> -
> > -------- This SF.net email is sponsored by Windows:
> >
> > Build for Windows Store.
> >
> > http://p.sf.net/sfu/windows-dev2dev
> > _______________________________________________
> > E1000-devel mailing list
> > [email protected]
> > https://lists.sourceforge.net/lists/listinfo/e1000-devel
> > To learn more about Intel&#174; Ethernet, visit
> > http://communities.intel.com/community/wired


------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:

Build for Windows Store.

http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel&#174; Ethernet, visit 
http://communities.intel.com/community/wired

Reply via email to