Asim
2008-Aug-08 03:31 UTC
[Xen-users] PCI Passthrough problems (Cannot access network but e1000 loads fine)
Hi, I''m running a PV DomU on CentOS 4. I was able to successfulyl hide a e1000 PCI card by appropriately modifying /etc/modprobe.conf. The device is not configured as a eth device in my dom0. I was also able to successfully load it in my domU and e1000 driver loads fine. ifconfig shows correct configuration of IP address. However, I''m not able to access the network. The network was accessible with the same IP using vif interface instead. Kindly help. Regards, Asim _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Andy Burns
2008-Aug-08 09:43 UTC
Re: [Xen-users] PCI Passthrough problems (Cannot access network but e1000 loads fine)
On 08/08/2008 04:31, Asim wrote:> I''m running a PV DomU on CentOS 4. I was able to successfulyl hide a > e1000 PCI card by appropriately modifying /etc/modprobe.conf. The > device is not configured as a eth device in my dom0. I was also able > to successfully load it in my domU and e1000 driver loads fine. > ifconfig shows correct configuration of IP address. > > However, I''m not able to access the network. The network was > accessible with the same IP using vif interface instead. Kindly help.Any ethernet or interrupt related messages in dmesg on dom0 or domU or from xm dmesg? Research the noirqdebug/pollirq/swiotlb settings and try some combinations _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Asim
2008-Aug-08 15:50 UTC
Re: [Xen-users] PCI Passthrough problems (Cannot access network but e1000 loads fine)
Hi, I do not get anything strange here. This is from the dmesg on domU [339497.233655] Intel(R) PRO/1000 Network Driver - version 7.1.9-k4-NAPI [339497.233657] Copyright (c) 1999-2006 Intel Corporation. [339497.234124] e1000: In __devinit. [339497.245661] PCI: Enabling device 0000:00:00.0 (0000 -> 0003) [339497.512872] e1000: 0000:00:00.0: e1000_probe: (PCI:33MHz:32-bit) 00:07:e9:39:07:e5 [339497.552859] e1000: eth0: e1000_probe: Intel(R) PRO/1000 Network Connection [339497.725307] Entered sys_init_module. [339497.730044] device-mapper: ioctl: 4.7.0-ioctl (2006-06-24) initialised: dm-devel@redhat.com [339498.000804] EXT3 FS on sda1, internal journal [339498.269163] Adding 65528k swap on /dev/sda2. Priority:-1 extents:1 across:65528k [339499.153997] e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps Full Duplex>From the prompt:[root@fedora_pristine ~]# ping 128.105.104.101 PING 128.105.104.101 (128.105.104.101) 56(84) bytes of data.>From 128.105.104.129 icmp_seq=1 Destination Host Unreachable >From 128.105.104.129 icmp_seq=2 Destination Host Unreachable >From 128.105.104.129 icmp_seq=3 Destination Host UnreachableIs there anything special I need to do for /etc/network or /etc/network-scripts/ifcfg-eth0 apart from what I did to get it working for vif interface. I used extra=''noirqdebug'' in my config file and my domU does not boot after : [340110.616654] Entered sys_init_module. [340110.621535] usbcore: registered new driver usbfs [340110.621572] usbcore: registered new driver hub I already have extra=''xcons=tty'' - does that get overidden? Please help. Regards, Asim On 8/8/08, Andy Burns <lists.xensource.com@adslpipe.co.uk> wrote:> On 08/08/2008 04:31, Asim wrote: > >> I''m running a PV DomU on CentOS 4. I was able to successfulyl hide a >> e1000 PCI card by appropriately modifying /etc/modprobe.conf. The >> device is not configured as a eth device in my dom0. I was also able >> to successfully load it in my domU and e1000 driver loads fine. >> ifconfig shows correct configuration of IP address. >> >> However, I''m not able to access the network. The network was >> accessible with the same IP using vif interface instead. Kindly help. > > Any ethernet or interrupt related messages in dmesg on dom0 or domU or > from xm dmesg? > > Research the noirqdebug/pollirq/swiotlb settings and try some combinations > > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Asim
2008-Aug-10 13:41 UTC
Re: [Xen-users] PCI Passthrough problems (Cannot access network but e1000 loads fine)
iptables is disabled on dom0. lspci shows the PCI device in dom0 but on domU it says no such command. Is there anything I need to add to mkinitrd to get this working or any lib to be copied? ifconfig tells me that the device acquires correct MAC and correct ipaddress. Thanks, Asim On Sun, Aug 10, 2008 at 12:16 AM, Devdutt Patnaik <xendevid@gmail.com> wrote:> Just my 2 cents, but are your iptables settings ok ? > Some other things to try : > > lspci - check to see that the device shows up fine. > ifconfig - verify that the device has the correct device mac address. > > Try "service network restart". > > If you''ve already verified all of these then it must be something else. > > -Devdutt. > > On Fri, Aug 8, 2008 at 11:50 AM, Asim <linkasim@gmail.com> wrote: >> >> Hi, >> >> I do not get anything strange here. This is from the dmesg on domU >> >> [339497.233655] Intel(R) PRO/1000 Network Driver - version 7.1.9-k4-NAPI >> [339497.233657] Copyright (c) 1999-2006 Intel Corporation. >> [339497.234124] e1000: In __devinit. >> [339497.245661] PCI: Enabling device 0000:00:00.0 (0000 -> 0003) >> [339497.512872] e1000: 0000:00:00.0: e1000_probe: (PCI:33MHz:32-bit) >> 00:07:e9:39:07:e5 >> [339497.552859] e1000: eth0: e1000_probe: Intel(R) PRO/1000 Network >> Connection >> [339497.725307] Entered sys_init_module. >> [339497.730044] device-mapper: ioctl: 4.7.0-ioctl (2006-06-24) >> initialised: dm-devel@redhat.com >> [339498.000804] EXT3 FS on sda1, internal journal >> [339498.269163] Adding 65528k swap on /dev/sda2. Priority:-1 >> extents:1 across:65528k >> [339499.153997] e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps Full >> Duplex >> >From the prompt: >> [root@fedora_pristine ~]# ping 128.105.104.101 >> PING 128.105.104.101 (128.105.104.101) 56(84) bytes of data. >> >From 128.105.104.129 icmp_seq=1 Destination Host Unreachable >> >From 128.105.104.129 icmp_seq=2 Destination Host Unreachable >> >From 128.105.104.129 icmp_seq=3 Destination Host Unreachable >> >> >> Is there anything special I need to do for /etc/network or >> /etc/network-scripts/ifcfg-eth0 apart from what I did to get it >> working for vif interface. >> >> I used extra=''noirqdebug'' in my config file and my domU does not boot >> after : >> [340110.616654] Entered sys_init_module. >> [340110.621535] usbcore: registered new driver usbfs >> [340110.621572] usbcore: registered new driver hub >> >> I already have extra=''xcons=tty'' - does that get overidden? >> >> Please help. >> >> Regards, >> Asim >> >> On 8/8/08, Andy Burns <lists.xensource.com@adslpipe.co.uk> wrote: >> > On 08/08/2008 04:31, Asim wrote: >> > >> >> I''m running a PV DomU on CentOS 4. I was able to successfulyl hide a >> >> e1000 PCI card by appropriately modifying /etc/modprobe.conf. The >> >> device is not configured as a eth device in my dom0. I was also able >> >> to successfully load it in my domU and e1000 driver loads fine. >> >> ifconfig shows correct configuration of IP address. >> >> >> >> However, I''m not able to access the network. The network was >> >> accessible with the same IP using vif interface instead. Kindly help. >> > >> > Any ethernet or interrupt related messages in dmesg on dom0 or domU or >> > from xm dmesg? >> > >> > Research the noirqdebug/pollirq/swiotlb settings and try some >> > combinations >> > >> > >> > >> > _______________________________________________ >> > Xen-users mailing list >> > Xen-users@lists.xensource.com >> > http://lists.xensource.com/xen-users >> > >> >> _______________________________________________ >> Xen-users mailing list >> Xen-users@lists.xensource.com >> http://lists.xensource.com/xen-users > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Asim
2008-Aug-10 19:26 UTC
Re: [Xen-users] PCI Passthrough problems (Cannot access network but e1000 loads fine)
I also found this message from xm demsg: (XEN) mm.c:645:d0 Non-privileged (178) attempt to map I/O space 00000000 Is this due to PCI passthrough. Please help. Regards, Asim On 8/10/08, Asim <linkasim@gmail.com> wrote:> iptables is disabled on dom0. > lspci shows the PCI device in dom0 but on domU it says no such > command. Is there anything I need to add to mkinitrd to get this > working or any lib to be copied? > > ifconfig tells me that the device acquires correct MAC and correct > ipaddress. > > Thanks, > Asim > > On Sun, Aug 10, 2008 at 12:16 AM, Devdutt Patnaik <xendevid@gmail.com> > wrote: >> Just my 2 cents, but are your iptables settings ok ? >> Some other things to try : >> >> lspci - check to see that the device shows up fine. >> ifconfig - verify that the device has the correct device mac address. >> >> Try "service network restart". >> >> If you''ve already verified all of these then it must be something else. >> >> -Devdutt. >> >> On Fri, Aug 8, 2008 at 11:50 AM, Asim <linkasim@gmail.com> wrote: >>> >>> Hi, >>> >>> I do not get anything strange here. This is from the dmesg on domU >>> >>> [339497.233655] Intel(R) PRO/1000 Network Driver - version 7.1.9-k4-NAPI >>> [339497.233657] Copyright (c) 1999-2006 Intel Corporation. >>> [339497.234124] e1000: In __devinit. >>> [339497.245661] PCI: Enabling device 0000:00:00.0 (0000 -> 0003) >>> [339497.512872] e1000: 0000:00:00.0: e1000_probe: (PCI:33MHz:32-bit) >>> 00:07:e9:39:07:e5 >>> [339497.552859] e1000: eth0: e1000_probe: Intel(R) PRO/1000 Network >>> Connection >>> [339497.725307] Entered sys_init_module. >>> [339497.730044] device-mapper: ioctl: 4.7.0-ioctl (2006-06-24) >>> initialised: dm-devel@redhat.com >>> [339498.000804] EXT3 FS on sda1, internal journal >>> [339498.269163] Adding 65528k swap on /dev/sda2. Priority:-1 >>> extents:1 across:65528k >>> [339499.153997] e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps Full >>> Duplex >>> >From the prompt: >>> [root@fedora_pristine ~]# ping 128.105.104.101 >>> PING 128.105.104.101 (128.105.104.101) 56(84) bytes of data. >>> >From 128.105.104.129 icmp_seq=1 Destination Host Unreachable >>> >From 128.105.104.129 icmp_seq=2 Destination Host Unreachable >>> >From 128.105.104.129 icmp_seq=3 Destination Host Unreachable >>> >>> >>> Is there anything special I need to do for /etc/network or >>> /etc/network-scripts/ifcfg-eth0 apart from what I did to get it >>> working for vif interface. >>> >>> I used extra=''noirqdebug'' in my config file and my domU does not boot >>> after : >>> [340110.616654] Entered sys_init_module. >>> [340110.621535] usbcore: registered new driver usbfs >>> [340110.621572] usbcore: registered new driver hub >>> >>> I already have extra=''xcons=tty'' - does that get overidden? >>> >>> Please help. >>> >>> Regards, >>> Asim >>> >>> On 8/8/08, Andy Burns <lists.xensource.com@adslpipe.co.uk> wrote: >>> > On 08/08/2008 04:31, Asim wrote: >>> > >>> >> I''m running a PV DomU on CentOS 4. I was able to successfulyl hide a >>> >> e1000 PCI card by appropriately modifying /etc/modprobe.conf. The >>> >> device is not configured as a eth device in my dom0. I was also able >>> >> to successfully load it in my domU and e1000 driver loads fine. >>> >> ifconfig shows correct configuration of IP address. >>> >> >>> >> However, I''m not able to access the network. The network was >>> >> accessible with the same IP using vif interface instead. Kindly help. >>> > >>> > Any ethernet or interrupt related messages in dmesg on dom0 or domU or >>> > from xm dmesg? >>> > >>> > Research the noirqdebug/pollirq/swiotlb settings and try some >>> > combinations >>> > >>> > >>> > >>> > _______________________________________________ >>> > Xen-users mailing list >>> > Xen-users@lists.xensource.com >>> > http://lists.xensource.com/xen-users >>> > >>> >>> _______________________________________________ >>> Xen-users mailing list >>> Xen-users@lists.xensource.com >>> http://lists.xensource.com/xen-users >> >> >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Asim
2008-Aug-11 16:11 UTC
Re: [Xen-users] PCI Passthrough problems (Cannot access network but e1000 loads fine)
Hello all, I browsed all archives of mailing list and saw that people use swiotlb=force to get it working. So i added in my dom0 swiotlb=256 noirqdebug and in my domU swiotlb=force. But this did not help me. Can anyone tell me what this does? Does it just simply enable a S/W IOMMU? My domU is still not working. I restored the vif line vif = [ ''mac=00:16:3E:6F:CF:77, bridge=eth0'' ] in my domU config file and it is able to ping. However, am not to sure if this is doing passthrough I/O and my belief is that it is doing bridged networking. Can anyone please help me. Regards, Asim On 8/10/08, Asim <linkasim@gmail.com> wrote:> I also found this message from xm demsg: > > (XEN) mm.c:645:d0 Non-privileged (178) attempt to map I/O space 00000000 > > Is this due to PCI passthrough. > > Please help. > > Regards, > Asim > > On 8/10/08, Asim <linkasim@gmail.com> wrote: >> iptables is disabled on dom0. >> lspci shows the PCI device in dom0 but on domU it says no such >> command. Is there anything I need to add to mkinitrd to get this >> working or any lib to be copied? >> >> ifconfig tells me that the device acquires correct MAC and correct >> ipaddress. >> >> Thanks, >> Asim >> >> On Sun, Aug 10, 2008 at 12:16 AM, Devdutt Patnaik <xendevid@gmail.com> >> wrote: >>> Just my 2 cents, but are your iptables settings ok ? >>> Some other things to try : >>> >>> lspci - check to see that the device shows up fine. >>> ifconfig - verify that the device has the correct device mac address. >>> >>> Try "service network restart". >>> >>> If you''ve already verified all of these then it must be something else. >>> >>> -Devdutt. >>> >>> On Fri, Aug 8, 2008 at 11:50 AM, Asim <linkasim@gmail.com> wrote: >>>> >>>> Hi, >>>> >>>> I do not get anything strange here. This is from the dmesg on domU >>>> >>>> [339497.233655] Intel(R) PRO/1000 Network Driver - version 7.1.9-k4-NAPI >>>> [339497.233657] Copyright (c) 1999-2006 Intel Corporation. >>>> [339497.234124] e1000: In __devinit. >>>> [339497.245661] PCI: Enabling device 0000:00:00.0 (0000 -> 0003) >>>> [339497.512872] e1000: 0000:00:00.0: e1000_probe: (PCI:33MHz:32-bit) >>>> 00:07:e9:39:07:e5 >>>> [339497.552859] e1000: eth0: e1000_probe: Intel(R) PRO/1000 Network >>>> Connection >>>> [339497.725307] Entered sys_init_module. >>>> [339497.730044] device-mapper: ioctl: 4.7.0-ioctl (2006-06-24) >>>> initialised: dm-devel@redhat.com >>>> [339498.000804] EXT3 FS on sda1, internal journal >>>> [339498.269163] Adding 65528k swap on /dev/sda2. Priority:-1 >>>> extents:1 across:65528k >>>> [339499.153997] e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps >>>> Full >>>> Duplex >>>> >From the prompt: >>>> [root@fedora_pristine ~]# ping 128.105.104.101 >>>> PING 128.105.104.101 (128.105.104.101) 56(84) bytes of data. >>>> >From 128.105.104.129 icmp_seq=1 Destination Host Unreachable >>>> >From 128.105.104.129 icmp_seq=2 Destination Host Unreachable >>>> >From 128.105.104.129 icmp_seq=3 Destination Host Unreachable >>>> >>>> >>>> Is there anything special I need to do for /etc/network or >>>> /etc/network-scripts/ifcfg-eth0 apart from what I did to get it >>>> working for vif interface. >>>> >>>> I used extra=''noirqdebug'' in my config file and my domU does not boot >>>> after : >>>> [340110.616654] Entered sys_init_module. >>>> [340110.621535] usbcore: registered new driver usbfs >>>> [340110.621572] usbcore: registered new driver hub >>>> >>>> I already have extra=''xcons=tty'' - does that get overidden? >>>> >>>> Please help. >>>> >>>> Regards, >>>> Asim >>>> >>>> On 8/8/08, Andy Burns <lists.xensource.com@adslpipe.co.uk> wrote: >>>> > On 08/08/2008 04:31, Asim wrote: >>>> > >>>> >> I''m running a PV DomU on CentOS 4. I was able to successfulyl hide a >>>> >> e1000 PCI card by appropriately modifying /etc/modprobe.conf. The >>>> >> device is not configured as a eth device in my dom0. I was also able >>>> >> to successfully load it in my domU and e1000 driver loads fine. >>>> >> ifconfig shows correct configuration of IP address. >>>> >> >>>> >> However, I''m not able to access the network. The network was >>>> >> accessible with the same IP using vif interface instead. Kindly help. >>>> > >>>> > Any ethernet or interrupt related messages in dmesg on dom0 or domU or >>>> > from xm dmesg? >>>> > >>>> > Research the noirqdebug/pollirq/swiotlb settings and try some >>>> > combinations >>>> > >>>> > >>>> > >>>> > _______________________________________________ >>>> > Xen-users mailing list >>>> > Xen-users@lists.xensource.com >>>> > http://lists.xensource.com/xen-users >>>> > >>>> >>>> _______________________________________________ >>>> Xen-users mailing list >>>> Xen-users@lists.xensource.com >>>> http://lists.xensource.com/xen-users >>> >>> >> >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users