From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jean Guyader Subject: Re: FLR support in xl tool stack Date: Fri, 28 Jan 2011 09:50:24 +0530 Message-ID: References: <987664A83D2D224EAE907B061CE93D53019438F9F8@orsmsx505.amr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: <987664A83D2D224EAE907B061CE93D53019438F9F8@orsmsx505.amr.corp.intel.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: "Kay, Allen M" Cc: xen-devel , Stefano Stabellini , Wilk , Jean Guyader , Ian Campbell , Thomas Horsten , Ross Philipson List-Id: xen-devel@lists.xenproject.org On 28 January 2011 08:38, Kay, Allen M wrote: > In xm tool stack, there is a function called do_FLR() located in python/x= en/util/pci.py that takes care of issuing FLR to passthrough PCI devices. = =A0This function is responsible for looking at the device and figure out th= e most appropriate way to reset the device (PCI/PCIe FLR, secondary bus res= et, etc...). > > In xl tool stack, I see there is a libxl_device_pci_reset() in libxl/libx= l_pci.c that suppose to do the similar thing. =A0It references /sys/bus/pci= /drivers/pciback/do_flr and tries to write something to it. =A0However, pci= back driver in 2.6.32.27 pvops kernel does not seem to have this do_flr fun= ctionality. > > Is this something being worked on but not checked in to pvops dom0 kernel= yet? > > If not, I would prefer implementing all of the do_FLR() functionaility in= libxl and not to depend on pciback driver given it is not part of the upst= ream kernel. =A0This will make using newer kernels such as 2.6.38 much easi= er. > > Comments? > On the newest kernels (I don't know from which version), the reset function has been moved under the pci device tree directly. So /sys/bus/pci/device//reset. Jean