James, At least one other person (Lta, included in this message) and I have had problems using passthrough pci devices and your GPLPV drivers at the same time. My symptoms are that SMB connections are totally unreliable (however, downloading over http seems to work well enough). For example, if I start a video or audio file, playing from the network, it will play the first few seconds fine, then the connection drops out. I can confirm that the problems don''t exist when not passing through any devices. I''m using xen-unstable c/s 24465, 3.2.1 dom0, win 7 64bit domU with an ATI 4770, USB controller, and ICE1712 based pci sound card passed through. I used the gplpv_Vista2008x64_0.11.0.308.msi drivers. I tried the -debug drivers, and I see a ton of output in qemu log, however, all that output seems to be at initialization. I do not see any additional output from your drivers after boot, including when I''m experiencing problems. I took a quick look at tcpdump output and it does seem the guest is ACKing the packets as they come in, so my first guess is that they''re getting lost somewhere between the driver and the OS network stack. I''ve never set up a build environment for these drivers so I haven''t tried adding any extra debug output. Have you looked into the (or a similar) problem before, or have a more verbose debug copy of the net driver you''ve used to diagnose similar problems before? I guess the real question is, do you have any idea where we should start looking? Thanks, Doug
On Wed 25. of January 2012 19:03:12 djmagee@mageenet.net wrote:> James, > At least one other person (Lta, included in this message) and I > have had problems using passthrough pci devices and your GPLPV drivers > at the same time. My symptoms are that SMB connections are totally > unreliable (however, downloading over http seems to work well enough). > For example, if I start a video or audio file, playing from the network, > it will play the first few seconds fine, then the connection drops out. > I can confirm that the problems don''t exist when not passing through any > devices. > > I''m using xen-unstable c/s 24465, 3.2.1 dom0, win 7 64bit domU > with an ATI 4770, USB controller, and ICE1712 based pci sound card > passed through. I used the gplpv_Vista2008x64_0.11.0.308.msi drivers. > > I guess the real question is, do you have any idea where we > should start looking?Hi, I had similar problem. I have several Win7 virtual machines with passed VGA where we play games with friends. Some games were not able to connect to LAN game when 0.11.0.308 were in place. Then I''ve installed 0.11.0.322 mentioned in http://old-list-archives.xen.org/archives/html/xen-devel/2011-09/msg01159.html and everything runs just fine. But they are not available to download anymore. -- Pavel Mateja
> > James, > At least one other person (Lta, included in this message) and I have > had problems using passthrough pci devices and your GPLPV drivers at the > same time. My symptoms are that SMB connections are totally unreliable > (however, downloading over http seems to work well enough). > For example, if I start a video or audio file, playing from the network, it will > play the first few seconds fine, then the connection drops out. > I can confirm that the problems don''t exist when not passing through any > devices. > > I''m using xen-unstable c/s 24465, 3.2.1 dom0, win 7 64bit domU with > an ATI 4770, USB controller, and ICE1712 based pci sound card passed > through. I used the gplpv_Vista2008x64_0.11.0.308.msi drivers. > > I tried the -debug drivers, and I see a ton of output in qemu log, > however, all that output seems to be at initialization. I do not see any > additional output from your drivers after boot, including when I''m > experiencing problems. > > I took a quick look at tcpdump output and it does seem the guest is > ACKing the packets as they come in, so my first guess is that they''re getting > lost somewhere between the driver and the OS network stack. > > I''ve never set up a build environment for these drivers so I haven''t > tried adding any extra debug output. Have you looked into the (or a similar) > problem before, or have a more verbose debug copy of the net driver you''ve > used to diagnose similar problems before? > > I guess the real question is, do you have any idea where we should > start looking? >I''d be first looking at interrupt sharing. Maybe GPLPV doesn''t like sharing interrupts with anything? Have a look in device manager and set the view as "Resources by type" and see if anything is sharing an interrupt with the "Xen PCI Device Driver". James
I checked, and no devices are sharing IRQs. I''ll try the windows2003/ndis5 drivers and see if that makes a difference. Of course, the base image for my win7 vms seems to have gotten corrupted, so I''ll have to build a new one.> -----Original Message----- > From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel- > bounces@lists.xensource.com] On Behalf Of James Harper > Sent: Wednesday, January 25, 2012 7:23 PM > To: djmagee@mageenet.net; lta@akr.fm > Cc: xen-devel@lists.xensource.com > Subject: Re: [Xen-devel] GPLPV and pci passthrough > > > > > James, > > At least one other person (Lta, included in this message) and I > have > > had problems using passthrough pci devices and your GPLPV drivers at > the > > same time. My symptoms are that SMB connections are totally > unreliable > > (however, downloading over http seems to work well enough). > > For example, if I start a video or audio file, playing from the > network, it will > > play the first few seconds fine, then the connection drops out. > > I can confirm that the problems don''t exist when not passing through > any > > devices. > > > > I''m using xen-unstable c/s 24465, 3.2.1 dom0, win 7 64bit domU > with > > an ATI 4770, USB controller, and ICE1712 based pci sound card passed > > through. I used the gplpv_Vista2008x64_0.11.0.308.msi drivers. > > > > I tried the -debug drivers, and I see a ton of output in qemu > log, > > however, all that output seems to be at initialization. I do notsee> any > > additional output from your drivers after boot, including when I''m > > experiencing problems. > > > > I took a quick look at tcpdump output and it does seem the guest > is > > ACKing the packets as they come in, so my first guess is thatthey''re> getting > > lost somewhere between the driver and the OS network stack. > > > > I''ve never set up a build environment for these drivers so I > haven''t > > tried adding any extra debug output. Have you looked into the (or a > similar) > > problem before, or have a more verbose debug copy of the net driver > you''ve > > used to diagnose similar problems before? > > > > I guess the real question is, do you have any idea where we > should > > start looking? > > > > I''d be first looking at interrupt sharing. Maybe GPLPV doesn''t like > sharing interrupts with anything? > > Have a look in device manager and set the view as "Resources by type" > and see if anything is sharing an interrupt with the "Xen PCI Device > Driver". > > James > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel
> -----Original Message----- > From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel- > bounces@lists.xensource.com] On Behalf Of djmagee@mageenet.net > Sent: Friday, January 27, 2012 2:03 PM > To: James Harper; lta@akr.fm > Cc: xen-devel@lists.xensource.com > Subject: Re: [Xen-devel] GPLPV and pci passthrough > > I checked, and no devices are sharing IRQs. I''ll try the > windows2003/ndis5 drivers and see if that makes a difference. Of > course, the base image for my win7 vms seems to have gotten corrupted, > so I''ll have to build a new one.Oops, sorry about the top post... I tried the Win2003 build and it works like a charm, and has native performance across my gbit network in at least one direction.> > > -----Original Message----- > > From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel- > > bounces@lists.xensource.com] On Behalf Of James Harper > > Sent: Wednesday, January 25, 2012 7:23 PM > > To: djmagee@mageenet.net; lta@akr.fm > > Cc: xen-devel@lists.xensource.com > > Subject: Re: [Xen-devel] GPLPV and pci passthrough > > > > > > > > James, > > > At least one other person (Lta, included in this message) and I > > have > > > had problems using passthrough pci devices and your GPLPV drivers > at > > the > > > same time. My symptoms are that SMB connections are totally > > unreliable > > > (however, downloading over http seems to work well enough). > > > For example, if I start a video or audio file, playing from the > > network, it will > > > play the first few seconds fine, then the connection drops out. > > > I can confirm that the problems don''t exist when not passing > through > > any > > > devices. > > > > > > I''m using xen-unstable c/s 24465, 3.2.1 dom0, win 7 64bit domU > > with > > > an ATI 4770, USB controller, and ICE1712 based pci sound card > passed > > > through. I used the gplpv_Vista2008x64_0.11.0.308.msi drivers. > > > > > > I tried the -debug drivers, and I see a ton of output in qemu > > log, > > > however, all that output seems to be at initialization. I do not > see > > any > > > additional output from your drivers after boot, including when I''m > > > experiencing problems. > > > > > > I took a quick look at tcpdump output and it does seem the guest > > is > > > ACKing the packets as they come in, so my first guess is that > they''re > > getting > > > lost somewhere between the driver and the OS network stack. > > > > > > I''ve never set up a build environment for these drivers so I > > haven''t > > > tried adding any extra debug output. Have you looked into the (or > a > > similar) > > > problem before, or have a more verbose debug copy of the netdriver> > you''ve > > > used to diagnose similar problems before? > > > > > > I guess the real question is, do you have any idea where we > > should > > > start looking? > > > > > > > I''d be first looking at interrupt sharing. Maybe GPLPV doesn''t like > > sharing interrupts with anything? > > > > Have a look in device manager and set the view as "Resources bytype"> > and see if anything is sharing an interrupt with the "Xen PCI Device > > Driver". > > > > James > > > > _______________________________________________ > > 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
> > > > I checked, and no devices are sharing IRQs. I''ll try the > > windows2003/ndis5 drivers and see if that makes a difference. Of > > course, the base image for my win7 vms seems to have gotten corrupted, > > so I''ll have to build a new one. > > Oops, sorry about the top post... > > I tried the Win2003 build and it works like a charm, and has native > performance across my gbit network in at least one direction. >Right... so if I understand the problem: . NDIS5 driver works fine under Windows 7 with PCI passthrough . NDIS6 driver works fine under Windows 7 with no PCI passthrough . NDIS6 driver does not work correctly with PCI passthrough Can you now try turning off all offloads in the DomU and see if starts working at some point? I''m at a loss to explain how this could happen though. James
On Sat, Jan 28, 2012 at 6:36 PM, James Harper <james.harper@bendigoit.com.au> wrote:>> > >> > I checked, and no devices are sharing IRQs. I''ll try the >> > windows2003/ndis5 drivers and see if that makes a difference. Of >> > course, the base image for my win7 vms seems to have gotten corrupted, >> > so I''ll have to build a new one. >> >> Oops, sorry about the top post... >> >> I tried the Win2003 build and it works like a charm, and has native >> performance across my gbit network in at least one direction. >> > > Right... so if I understand the problem: > > . NDIS5 driver works fine under Windows 7 with PCI passthrough > . NDIS6 driver works fine under Windows 7 with no PCI passthrough > . NDIS6 driver does not work correctly with PCI passthrough > > Can you now try turning off all offloads in the DomU and see if starts working at some point? > > I''m at a loss to explain how this could happen though. > > James > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-develI have had weird network problems as well using the 308 drivers. For me the 356 drivers from the hg repository do work fine, so something between 308 and 356 fixed some network bugs. Perhaps put a test build on the website for him to try? Roderick
> > > > I tried the Win2003 build and it works like a charm, and has native > > performance across my gbit network in at least one direction. > > > > Right... so if I understand the problem: > > . NDIS5 driver works fine under Windows 7 with PCI passthrough > . NDIS6 driver works fine under Windows 7 with no PCI passthrough > . NDIS6 driver does not work correctly with PCI passthroughCorrect.> > Can you now try turning off all offloads in the DomU and see if starts > working at some point?I disabled CSum offload, Large Send Offload, and scatter gather. No help. I also set HKLM\System\CurrentControlSet\services\tcpip\Parameters\DisableTaskOfflo ad=1. Also no help.> > I''m at a loss to explain how this could happen though. >If I think of anything else I''ll try it. Would it help to try a 32bit Windows 7 instance and see if there''s any difference?
> I have had weird network problems as well using the 308 drivers. For > me the 356 drivers from the hg repository do work fine, so something > between 308 and 356 fixed some network bugs. Perhaps put a test build > on the website for him to try? >I already have the Windows DDK installed on a couple of machines, but the link to the hg repo at http://wiki.xen.org/wiki/Xen_Windows_GplPv/Building seems to be broken. Where''s the current repo housed?
On Mon, Jan 30, 2012 at 7:22 PM, <djmagee@mageenet.net> wrote:>> I have had weird network problems as well using the 308 drivers. For >> me the 356 drivers from the hg repository do work fine, so something >> between 308 and 356 fixed some network bugs. Perhaps put a test build >> on the website for him to try? >> > > I already have the Windows DDK installed on a couple of machines, but > the link to the hg repo at > http://wiki.xen.org/wiki/Xen_Windows_GplPv/Building seems to be broken. > Where''s the current repo housed?You can find it here: http://xenbits.xen.org/ext/win-pvdrivers/ Look at the README file of the source tree for additional build instructions. Roderick
> On Mon, Jan 30, 2012 at 7:22 PM, <djmagee@mageenet.net> wrote: > >> I have had weird network problems as well using the 308 drivers.For> >> me the 356 drivers from the hg repository do work fine, sosomething> >> between 308 and 356 fixed some network bugs. Perhaps put a test > build > >> on the website for him to try? > >> > > > > I already have the Windows DDK installed on a couple of machines,but> > the link to the hg repo at > > http://wiki.xen.org/wiki/Xen_Windows_GplPv/Building seems to be > broken. > > Where''s the current repo housed? > > You can find it here: > http://xenbits.xen.org/ext/win-pvdrivers/ > > Look at the README file of the source tree for additional build > instructions. > > RoderickThanks, that was easy enough. I built the latest code available from the hg repo and I have the same symptoms. I''ll dig around and see if I can find a good place to dump some debug output and track down the problem. Doug