From mboxrd@z Thu Jan 1 00:00:00 1970 From: Don Dutile Subject: Re: Expose ltr/obff interface by sysfs Date: Fri, 06 Apr 2012 17:20:21 -0400 Message-ID: <4F7F5E15.3040808@redhat.com> References: <403610A45A2B5242BD291EDAE8B37D300FD0A434@SHSMSX102.ccr.corp.intel.com> <20120406082649.68fe5842@jbarnes-desktop> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: "Hao, Xudong" , "linux-pci@vger.kernel.org" , "netdev@vger.kernel.org" , "e1000-devel@lists.sourceforge.net" To: Jesse Barnes Return-path: In-Reply-To: <20120406082649.68fe5842@jbarnes-desktop> Sender: linux-pci-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On 04/06/2012 11:26 AM, Jesse Barnes wrote: > On Fri, 6 Apr 2012 02:43:59 +0000 > "Hao, Xudong" wrote: > >> Hi, >> >> I'm working on virtualization Xen/KVM. I saw there are ltr/obff enabling/disabling function in pci.c, but no called till now. I want to know if anybody(driver developer) are working for using it? Can driver change the LTR latency value dynamically? > > I believe the value is writable, but I'd expect some devices to > misbehave if the value were programmed too high. Performance would > also suffer if the value were set too high, at least for IOPS sensitive > devices. > >> /* >> LTR(Latency tolerance reporting) allows devices to send messages to the root complex indicating their latency tolerance for snooped& unsnooped memory transactions. >> OBFF (optimized buffer flush/fill), where supported, can help improve energy efficiency by giving devices information about when interrupts and other activity will have a reduced power impact. >> */ >> >> One way to control ltr/obff is used by driver, however, I'm considering that in virtualization, how guest OS driver control them. I have an idea that expose an inode interface by sysfs, like "reset" inode implemented in pci-sysfs.c, so that system user/administrator can enable/disable ltr/obff or set latency value on userspace, but not limited on driver. Comments? > > Given how device specific these extensions are, I'd expect you'd need > to know about each specific device anyway, which is why I think the > control belongs in the driver. I don't see why you'd need to > enable/disable/change these functions when assigning a device from one > guest to another... > and looking at this code last week, I noticed the ltr, obff & ido functions should be doing pcie_cap_has_*() checks before accessing registers that may not exist, as is done in pci_[save,restore]_pcie_state() ... in drivers/pci/pci.c: #define pcie_cap_has_devctl2(type, flags) \ ((flags & PCI_EXP_FLAGS_VERS) > 1) #define pcie_cap_has_lnkctl2(type, flags) \ ((flags & PCI_EXP_FLAGS_VERS) > 1) #define pcie_cap_has_sltctl2(type, flags) \ ((flags & PCI_EXP_FLAGS_VERS) > 1)