Hi, Is there a plan to enable the nested PCI bridge support of VT-d? Currently, if there is nested PCI bridge, a message will shown saying it''s not supported. And if passing the card on the slot on a nested PCI bridge, it seems some unexpected problem may occur. I encounter this case when trying to pass through a PCI card to a RHEL3 HVM on HP xw8600 workstation. The (only) PCI slot in xw8600 is in a nested PCI bridge, according to lspci -t. If pass the card on this PCI slot into a HVM, the system just hangs, I guess because some problem occurs and causes the LSI SCSI controller on the same PCI bridge not work correctly any more. On the same system, the pass through of another PCI device (a on-board PCI Express NIC) is working OK. Thanks, Jimmy Jin _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Hi Jimmy, All devices behind PCIe-to-PCI bridge has to be assigned to the same domain. Supporting nested PCI bridge is a little bit complicated, and even infeasible in some cases. I think it makes a little sense. It is meaningful to make an interface for users to know which devices are assignable with VT-d, and hint them to assign correctly. Randy (Weidong) Jimmy Jin wrote:> Hi, > > Is there a plan to enable the nested PCI bridge support of VT-d? > Currently, if there is nested PCI bridge, a message will shown saying > it''s not supported. And if passing the card on the slot on a nested > PCI bridge, it seems some unexpected problem may occur. > > I encounter this case when trying to pass through a PCI card to a > RHEL3 HVM on HP xw8600 workstation. The (only) PCI slot in xw8600 is > in a nested PCI bridge, according to lspci -t. If pass the card on > this PCI slot into a HVM, the system just hangs, I guess because some > problem occurs and causes the LSI SCSI controller on the same PCI > bridge not work correctly any more. On the same system, the pass > through of another PCI device (a on-board PCI Express NIC) is working > OK. > > Thanks, > Jimmy Jin > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Hello, Randy,
Thank you very much for your reply.
Just like to double confirm my understanding.
In my case the lspci -t -v output is like following:
-[0000:00]-+-00.0 Intel Corporation Memory Controller Hub
+-01.0-[0000:80]--
+-03.0-[0000:a0]--
+-05.0-[0000:60]----00.0 nVidia Corporation Quadro NVS 290
+-09.0-[0000:10-40]--+-00.0-[0000:1e-40]--+-00.0-[0000:20]--
| | \-01.0-[0000:40]--
| \-00.3-[0000:11]--+-06.0 LSI Logic /
Symbios Logic SAS1068 PCI-X Fusion-MPT SAS
| \-09.0 ADMtek NC100
Network Everywhere Fast Ethernet 10/100
...
I''m trying to pass through the ADMtek NC100 NIC to RHEL 3.7 HVM domU.
So according to your explanation, I must also pass the LSI Logic SAS
controller to the same domU, right? Otherwise the problem I
encountered with the SAS controller will occur, right?
If it''s correct, it seems the only solution on this workstation is to
have SATA HD instead of SAS. :-(
Thanks,
Jimmy Jin
On Wed, May 28, 2008 at 3:17 PM, Han, Weidong <weidong.han@intel.com>
wrote:> Hi Jimmy,
>
> All devices behind PCIe-to-PCI bridge has to be assigned to the same
> domain.
>
> Supporting nested PCI bridge is a little bit complicated, and even
> infeasible in some cases. I think it makes a little sense. It is
> meaningful to make an interface for users to know which devices are
> assignable with VT-d, and hint them to assign correctly.
>
> Randy (Weidong)
>
>
> Jimmy Jin wrote:
>> Hi,
>>
>> Is there a plan to enable the nested PCI bridge support of VT-d?
>> Currently, if there is nested PCI bridge, a message will shown saying
>> it''s not supported. And if passing the card on the slot on a
nested
>> PCI bridge, it seems some unexpected problem may occur.
>>
>> I encounter this case when trying to pass through a PCI card to a
>> RHEL3 HVM on HP xw8600 workstation. The (only) PCI slot in xw8600 is
>> in a nested PCI bridge, according to lspci -t. If pass the card on
>> this PCI slot into a HVM, the system just hangs, I guess because some
>> problem occurs and causes the LSI SCSI controller on the same PCI
>> bridge not work correctly any more. On the same system, the pass
>> through of another PCI device (a on-board PCI Express NIC) is working
>> OK.
>>
>> Thanks,
>> Jimmy Jin
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xensource.com
>> http://lists.xensource.com/xen-devel
>
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Jimmy Jin wrote:> Hello, Randy, > > Thank you very much for your reply. > > Just like to double confirm my understanding. > > In my case the lspci -t -v output is like following: > > -[0000:00]-+-00.0 Intel Corporation Memory Controller Hub > +-01.0-[0000:80]-- > +-03.0-[0000:a0]-- > +-05.0-[0000:60]----00.0 nVidia Corporation Quadro NVS 290 > > > > +-09.0-[0000:10-40]--+-00.0-[0000:1e-40]--+-00.0-[0000:20]-- | > | \-01.0-[0000:40]-- | > \-00.3-[0000:11]--+-06.0 LSI Logic / > Symbios Logic SAS1068 PCI-X Fusion-MPT SAS > | \-09.0 ADMtek NC100 > Network Everywhere Fast Ethernet 10/100 > ... > > I''m trying to pass through the ADMtek NC100 NIC to RHEL 3.7 HVM domU. > So according to your explanation, I must also pass the LSI Logic SAS > controller to the same domU, right? Otherwise the problem I > encountered with the SAS controller will occur, right?yes, you should assign all the devices under the same PCIe-to-PCI bridge to the same domain. Randy (Weidong)> > If it''s correct, it seems the only solution on this workstation is to > have SATA HD instead of SAS. :-( > > Thanks, > Jimmy Jin > > On Wed, May 28, 2008 at 3:17 PM, Han, Weidong <weidong.han@intel.com> > wrote: >> Hi Jimmy, >> >> All devices behind PCIe-to-PCI bridge has to be assigned to the same >> domain. >> >> Supporting nested PCI bridge is a little bit complicated, and even >> infeasible in some cases. I think it makes a little sense. It is >> meaningful to make an interface for users to know which devices are >> assignable with VT-d, and hint them to assign correctly. >> >> Randy (Weidong) >> >> >> Jimmy Jin wrote: >>> Hi, >>> >>> Is there a plan to enable the nested PCI bridge support of VT-d? >>> Currently, if there is nested PCI bridge, a message will shown >>> saying it''s not supported. And if passing the card on the slot on a >>> nested PCI bridge, it seems some unexpected problem may occur. >>> >>> I encounter this case when trying to pass through a PCI card to a >>> RHEL3 HVM on HP xw8600 workstation. The (only) PCI slot in xw8600 is >>> in a nested PCI bridge, according to lspci -t. If pass the card on >>> this PCI slot into a HVM, the system just hangs, I guess because >>> some problem occurs and causes the LSI SCSI controller on the same >>> PCI bridge not work correctly any more. On the same system, the pass >>> through of another PCI device (a on-board PCI Express NIC) is >>> working OK. >>> >>> Thanks, >>> Jimmy Jin >>> >>> _______________________________________________ >>> Xen-devel mailing list >>> Xen-devel@lists.xensource.com >>> http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Hello, Randy, After changing the harddrive of the workstation with a SATA one, so that I can avoid the LSI SAS controller on the same PCI-e-to-PCI bridge, now it works. Thanks, Jimmy Jin On Thu, May 29, 2008 at 5:18 PM, Han, Weidong <weidong.han@intel.com> wrote:> Jimmy Jin wrote: >> Hello, Randy, >> >> Thank you very much for your reply. >> >> Just like to double confirm my understanding. >> >> In my case the lspci -t -v output is like following: >> >> -[0000:00]-+-00.0 Intel Corporation Memory Controller Hub >> +-01.0-[0000:80]-- >> +-03.0-[0000:a0]-- >> +-05.0-[0000:60]----00.0 nVidia Corporation Quadro NVS 290 >> >> >> >> +-09.0-[0000:10-40]--+-00.0-[0000:1e-40]--+-00.0-[0000:20]-- | >> | \-01.0-[0000:40]-- | >> \-00.3-[0000:11]--+-06.0 LSI Logic / >> Symbios Logic SAS1068 PCI-X Fusion-MPT SAS >> | \-09.0 ADMtek NC100 >> Network Everywhere Fast Ethernet 10/100 >> ... >> >> I''m trying to pass through the ADMtek NC100 NIC to RHEL 3.7 HVM domU. >> So according to your explanation, I must also pass the LSI Logic SAS >> controller to the same domU, right? Otherwise the problem I >> encountered with the SAS controller will occur, right? > > yes, you should assign all the devices under the same PCIe-to-PCI bridge > to the same domain. > > Randy (Weidong) > >> >> If it''s correct, it seems the only solution on this workstation is to >> have SATA HD instead of SAS. :-( >> >> Thanks, >> Jimmy Jin >> >> On Wed, May 28, 2008 at 3:17 PM, Han, Weidong <weidong.han@intel.com> >> wrote: >>> Hi Jimmy, >>> >>> All devices behind PCIe-to-PCI bridge has to be assigned to the same >>> domain. >>> >>> Supporting nested PCI bridge is a little bit complicated, and even >>> infeasible in some cases. I think it makes a little sense. It is >>> meaningful to make an interface for users to know which devices are >>> assignable with VT-d, and hint them to assign correctly. >>> >>> Randy (Weidong) >>> >>> >>> Jimmy Jin wrote: >>>> Hi, >>>> >>>> Is there a plan to enable the nested PCI bridge support of VT-d? >>>> Currently, if there is nested PCI bridge, a message will shown >>>> saying it''s not supported. And if passing the card on the slot on a >>>> nested PCI bridge, it seems some unexpected problem may occur. >>>> >>>> I encounter this case when trying to pass through a PCI card to a >>>> RHEL3 HVM on HP xw8600 workstation. The (only) PCI slot in xw8600 is >>>> in a nested PCI bridge, according to lspci -t. If pass the card on >>>> this PCI slot into a HVM, the system just hangs, I guess because >>>> some problem occurs and causes the LSI SCSI controller on the same >>>> PCI bridge not work correctly any more. On the same system, the pass >>>> through of another PCI device (a on-board PCI Express NIC) is >>>> working OK. >>>> >>>> Thanks, >>>> Jimmy Jin >>>> >>>> _______________________________________________ >>>> Xen-devel mailing list >>>> Xen-devel@lists.xensource.com >>>> http://lists.xensource.com/xen-devel > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel