From mboxrd@z Thu Jan 1 00:00:00 1970 From: Konrad Rzeszutek Wilk Subject: Re: how to PCI-passthrough 1 card to a DomU when another @ Dom0 uses the same driver? Date: Thu, 3 Feb 2011 11:26:52 -0500 Message-ID: <20110203162652.GA3712@dumpdata.com> References: <1296623011.3053.1418495679@webmail.messagingengine.com> <20110202202705.GD19684@dumpdata.com> <1296683031.21707.1418639239@webmail.messagingengine.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <1296683031.21707.1418639239@webmail.messagingengine.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: dev001 Cc: xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org > i'm honestly not sure what to check, here. what, specifically, am i > looking for? readlink //sys/bus/pci/devices/0000:04:00.0/driver ../../../../bus/pci/drivers/pciback I want you to do for 02:00.0 and see if it points to pciback before you launch any guests. > > > > i suspect this has to do with the driver not being unloadable at the > > > time of device seizure, as it's in use. > > > > Well, the pciback.hide=(..) during bootup should have taken care of that. > > > > Oh, unless the pciback is a _module_ at which point it would not do it. > > on opensuse, iiuc it is a module > > grep CONFIG_XEN_PCIDEV_BACKEND /boot/config-2.6.34.8-7-xen > CONFIG_XEN_PCIDEV_BACKEND=m <-------------------- Pfff.. sucks. > CONFIG_XEN_PCIDEV_BACKEND_VPCI=y > # CONFIG_XEN_PCIDEV_BACKEND_PASS is not set > # CONFIG_XEN_PCIDEV_BACKEND_SLOT is not set > # CONFIG_XEN_PCIDEV_BACKEND_CONTROLLER is not set > .. snip.. > and, since it's compiled as a module, at > > http://wiki.xensource.com/xenwiki/Assign_hardware_to_DomU_with_PCIBack_as_module > > i "followed the rules" for method 2 ... > Yup. > is that not correct? > > > What happens if you 'rmmod sky2' or blacklist the sky2 driver? > > perhaps i'm being thick -- what's that tell us? i've already shown that > replacing the 'first' r8169-based card with something else allows BOTH > to-be-passed-thru pci-cards to be seized correctly. which seems to me It eliminates the possiblity of the sky2 driver having ownership of the 02:00.0. It should be assigned to the pciback driver when you launch your guest. > to indicate that i'm set up correctly. I want to eliminate the possiblity of the 02:00.0 being owned by somebody else. If it is owned by pciback, then yes it looks as if there is a bug in either the xm tools or the pciback driver when it comes to seizing the BDF that have the same PCI vendor:model