tag:linuxplumbersconf.org,2005:/2015/ocw/proposalsLinux Plumbers Conference: Presentation Proposals2015-08-21T13:16:10-07:00tag:linuxplumbersconf.org,2005:Proposal/34592015-08-21T13:16:10-07:002015-08-21T13:19:54-07:00Plumbers PostMortem Discussion2015-08-21T20:16:10Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Come and give Audience feedback about what worked and what didn't work in Plumbers this year. Feedback in person is gratefully received but feedback to mailto:contact@list.linuxplumbersconf.org is also gratefully accepted</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Come and give Audience feedback about what worked and what didn’t work in Plumbers this year.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>James E.J. Bottomley</b></p><div/><p>James Bottomley is <span class="caps">CTO</span> of Server Virtualisation at Odin and Linux Kernel maintainer of the <span class="caps">SCSI</span> subsystem, PA-<span class="caps">RISC</span> Linux and the 53c700 set of drivers. He has made contributions in the areas of x86 architecture and <span class="caps">SMP</span>, filesystems, storage and memory management and coherency. He is currently a Director on the Board of the Linux Foundation and Chair of its Technical Advisory Board. He was born and<br />
grew up in the United Kingdom. He went to university t Cambridge in 1985 for both his undergraduate and doctoral degrees. He joined AT&T Bell labs in 1995 to work on Distributed Lock Manager technology for clustering. In 1997 he moved to the LifeKeeper High Availability project. In 2000 he helped found SteelEye Technology, Inc as Software Architect and later as Vice President and <span class="caps">CTO</span>. He joined Novell in 2008 as a Distinguished Engineer at Novell’s <span class="caps">SUSE</span> Labs and arallels (now Odin) in 2011.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34532015-08-18T11:54:54-07:002015-08-20T13:38:55-07:00BOF: Distributed file systems2015-08-18T18:54:54Z<div><dl><dt><b>One Line Summary:</b></dt><dd>In-kernel NFS, NFS/Ganesha, Samba, CIFS, Ceph and more</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Gathering of those interested in discussing their top concerns about distributed filesystem implementations in Linux. Performance, security, user management, new networking technologies, and so on.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Chuck Lever</b></p><div/><p>Linux <span class="caps">NFS</span> hacker</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34472015-08-17T11:09:46-07:002015-08-18T22:25:55-07:00BOF: Potential Userspace Filesystem APIs2015-08-17T18:09:46Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Let's get together to discuss potential userspace APIs for advanced FS & storage systems that we could add to the kernel!</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The objective of this <span class="caps">BOF</span> session is to discuss what kinds of new FS & storage APIs could be useful additions for userspace programs. Things that we can talk about include a more standard syscall for accessing accelerated copy routines (filesystem reflink, the <span class="caps">SCSI</span> copy offloading thing); the proposed <span class="caps">XFS</span> configurable error handling interface; and anything else that application programmers might want to discuss with kernel engineers.</p>
<p>(There is a separate <span class="caps">SMR</span> <span class="caps">BOF</span>.)</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Darrick Wong</b></p><div/><p>Darrick has been with Oracle’s Linux development team for the past three years. In that time he has worked on many areas of the kernel, most notably environmental sensors, storage drivers, energy management, ext4, and most recently <span class="caps">XFS</span>.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34412015-08-13T21:06:34-07:002015-08-25T14:37:56-07:00Chat With The dtc Maintainers (continued)2015-08-14T04:06:34Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A discussion with the dtc maintainers. (No slides presented.)</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>This session is an opportunity to ask questions of the dtc maintainers or listen to their thoughts on dtc related topics.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Frank Rowand</b></p><div/><p>Frank has meddled in the internals of several proprietary operating systems, but has been loyal to the Linux kernel since 1999. He has worked in many areas of technology, including performance, networking, platform support, drivers, real-time, and embedded. He is currently employed by Sony Mobile Communications.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34352015-08-13T11:31:08-07:002015-08-20T08:35:03-07:00State of Wayland2015-08-13T18:31:08Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Quick overview of the development status of Wayland and Weston.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The core Wayland protocol is quite mature, and ready for desktop environments to adopt. That said, we do have a couple race conditions and a rendering issue relating to damage with rotated surfaces.</p>
<p>Weston serves as a reference implementation and as a testbed for new technology, some of which will eventually migrate to Wayland. We’ll take a look at some of the things being worked on in Weston and things coming soon.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Bryce Harrington</b></p><div/><p>Bryce Harrington is a Senior Open Source Developer at Samsung Research America focusing on Open Source Graphics. Prior to Samsung, he led Canonical, Ltd.’s Ubuntu X.org team, and focused on stabilization of the graphics and input infrastructures for the Ubuntu distribution. Bryce began his career in the aerospace industry as a spacecraft propulsion engineer at The Aerospace Corporation, Hughes Space and Communications and <span class="caps">TRW</span>. Later, he joined the Open Source Development Labs as a Senior Performance Engineer working on NFSv4 testing and development of automated test systems. He is a founder and developer of the Inkscape project and serves as Chairman of the Inkscape Board. Bryce has a BS-AE from <span class="caps">USC</span> and an MS-AE from Caltech.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34292015-07-31T13:42:09-07:002015-08-03T21:01:32-07:00Legal, Licensing and Enforcement Workshop for Plumbers2015-07-31T20:42:09Z<div><dl><dt><b>One Line Summary:</b></dt><dd>What should you know about Licenses and how should you use them. Join Karen Sandler, Executive Director of Conservancy and Denver Gingerich, License Compliance Engineer of Conservancy, for an in-depth Q&A and discussion session.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>What should you know about Licenses and how should you use them. Join Karen Sandler, Executive Director of Conservancy and Denver Gingerich, License Compliance Engineer of Conservancy, for an in-depth Q&A and discussion session.</p>
<p>You can turn up with any legal or licensing questions you may have about Plumbing code and Karen will endeavour to answer them, though she is not your lawyer and this won’t be legal advice! Denver will provide insight into the nuts and bolts of compliance.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Karen Sandler</b></p><div/><p>Karen M. Sandler is Executive Director of the Software Freedom Conservancy, the nonprofit home of dozens of essential free software projects, including the <span class="caps">GPL</span> Compliance Project for Linux Kernel Developers. She is known for her advocacy for free and open source software, particularly in relation to the software on medical devices. She was previously the Executive Director of the <span class="caps">GNOME</span> Foundation. Karen co-organizes Outreachy (formerly Outreach Program for Women). She received an O’Reilly Open Source Award and is co-host of the oggcast “Free and in Freedom”. Karen is a lawyer, admitted to practice in the state of New York.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34232015-07-28T22:24:42-07:002015-07-28T22:24:42-07:00MD RAID 5/6 Caching Layer2015-07-29T05:24:42Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discuss using SSD/NVRAM to cache MD RAID 5/6 data</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Linux MD <span class="caps">RAID</span> 5/6 is widely adopted, but it has limitations. Partial stripe write involves read-modify-write, which is a severe performance bottleneck. MD <span class="caps">RAID</span> 5/6 also has a reliability issue, called write hole. Adding a caching layer to MD <span class="caps">RAID</span> 5/6 kills two birds with one stone. Plus, a caching layer makes <span class="caps">RAID</span> 5/6 resynchronization after a unclean shutdown unnecessary. The caching layer can reside in a <span class="caps">SSD</span> or <span class="caps">NDRAM</span>. This talk will discuss the status/issues in this area.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Shaohua Li</b></p><div/><p>Linux Kernel Developer in Facebook</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34172015-07-27T03:41:45-07:002015-08-14T10:32:15-07:00Integrating KDBus in Android2015-07-27T10:41:45Z<div><dl><dt><b>One Line Summary:</b></dt><dd>An experiment on using KDBus as a potential replacement for Binder in Android. </dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Over the course of three months we have experimented with integrating KDBus in Android. More specifically, considering it as a potential replacement for Binder in the future. KDBus is being considered for upstream inclusion which makes now the perfect time for us to experiment with it.</p>
<p>We will talk about how Android currently performs <span class="caps">IPC</span> with the Binder <span class="caps">API</span>. And then go into a little more details of how this <span class="caps">API</span> is implemented on top of the binder kernel driver. We will also go through KDBus’ capabilities in order to see how it could satisfy Binder’s needs and model.</p>
<p>Finally, we will see a possible approach: Implementing a subset of the Binder <span class="caps">API</span> on top <span class="caps">KDBUS</span>. This will allow us to show how Binder’s synchronous nature can be implemented and present a way to move forward with this work.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Pierre Langlois</b></p><div/><p>I am a Graduate software engineer at <span class="caps">ARM</span>. I have been involved in various project over the past couple of years including Android, Google-V8 and the <span class="caps">GNU</span> tools.</p>
<p>Originally from Franche-Comté, France, I graduated in Embedded systems and Computer science engineering at the university of Burgundy, <span class="caps">ESIREM</span> in Dijon.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34112015-07-25T12:44:28-07:002015-07-25T12:51:36-07:00State of Staging2015-07-25T19:44:28Z<div><dl><dt><b>One Line Summary:</b></dt><dd>State of the Android staging tree code</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>General discussion about what the state is of the code in the drivers/staging/android/ portion of the kernel, what needs to be deleted, what needs to be merged to the main portion of the kernel, and what needs to be added to it from external repositories.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Greg Kroah-Hartman</b></p><div/><p>Greg is a Linux kernel developer.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/34052015-07-24T19:11:27-07:002015-08-20T18:48:10-07:00Adapting Android for Ara2015-07-25T02:11:27Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion of how Android is being modified to work for Project Ara.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Android has existed for about a decade now and, during that time, it has mostly run on hardware that was fixed at design time. As such, Android images typically contain support for a very specific set of hardware that does not vary at runtime. Project Ara revisits this classic mobile design by introducing a modular handset that allows its user to dynamically change almost everything on his device. To work on such a platform, Android needs to be modified in several ways. This talk/discussion will cover some of the ongoing work in this area.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Karim Yaghmour</b></p><div/><p>Karim Yaghmour is <span class="caps">CEO</span> of Opersys, which specializes in Embedded Android and Embedded Linux, and best defines himself as part serial entrepreneur, part unrepentant geek. He’s the author of O’Reilly’s Embedded Android, the first book to cover Android’s internals, and Building Embedded Linux Systems, which sold tens of thousands of copies worldwide and has been translated into several different languages. Karim pioneered the world of Linux tracing by introducing the Linux Trace Toolkit (<span class="caps">LTT</span>) in the late ’90s. Other open source contributions include relayfs and Adeos. Karim has presented and published as part of a number of peer-reviewed scientific conferences, magazines and online publications, including Usenix, the Linux Kernel Summit, the Embedded Linux Conference, the Android Builders Summit, and the Real-Time Linux Workshop.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33992015-07-24T12:02:33-07:002015-09-09T17:19:59-07:00Improving big.LITTLE on Android2015-07-24T19:02:33Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Status of heterogeneous CPU support in Android and new features included in Android M to improve it.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Discussion of big.<span class="caps">LITTLE</span> processors, areas we see as a problem in Android, and improvements we’ve made to Android to make big.<span class="caps">LITTLE</span> more efficient. Also, discussion of future improvements we could make in Android to further improve scheduling.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Tim Murray</b></p><div/><p>Tim is a software engineer on the Android platform team at Google. His primary responsibilities include performance tuning and measurement.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33932015-07-23T14:30:14-07:002015-08-12T04:42:55-07:00BOF: Bittern Disk Cache2015-07-23T21:30:14Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Bittern Disk Cache: replacing Hardware RAID with Persistent Memory and software RAID stack</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>In this presentation, Fio Cattaneo, will talk about how Twitter Cache takes advantage of Persistent Memory technology (<span class="caps">NVDIMM</span> and PCIe-<span class="caps">NVRAM</span>) and software <span class="caps">RAID</span> to replace the functionality of HW <span class="caps">RAID</span>. Its main advantages are lower costs, the use of commodity hardware and open source stack versus proprietary HW <span class="caps">RAID</span> solutions. Twitter Cache is tuned towards optimizing the use of persistent memory which is smaller but faster than <span class="caps">SSD</span> caching solutions. Eviction algorithms are tuned towards a write buffer to best utilize the smaller size. Cache access uses a low overhead <span class="caps">API</span> as opposed to block store. Metadata is kept in memory for fast access. Because of its low latency, it also allows for caching of <span class="caps">SSD</span> storage. Twitter Cache is being actively developed and tested on Twitter workloads.</p>
<p>Bittern was already presented at Linuxcon 2014 <a href="http://events.linuxfoundation.org/sites/events/files/slides/Bittern_Linuxcon_2014_v0.12.pdf" rel="nofollow">http://events.linuxfoundation.org/sites/events/files/slides/Bittern_Linuxcon_2014_v0.12.pdf</a> . Bittern is now being deployed within Twitter for transactional workloads.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Fiorenzo Cattaneo</b></p><div/><p>Fio Cattaneo has been working in the industry for 25 years and his main interests are in kernels, operating systems and distributed systems. His first kernel work was at Olivetti S.p.A., help porting BSD4.2 on a Motorola architecture, and subsequently develop multiprocessor support in the early days of <span class="caps">SMP</span> technology. More recently he worked on distributed systems and storage at Pyramid Technology and Bluearc. He worked at Microsoft in the clustering group, and at Amazon <span class="caps">AWS</span> on distributed systems and storage technologies. He joined Twitter in early 2013 and works for the OS Team on the Twitter Cache project.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33872015-07-20T13:02:13-07:002015-08-20T13:39:21-07:00Android, partitions and customization2015-07-20T20:02:13Z<div><dl><dt><b>One Line Summary:</b></dt><dd>How the Android project is trying to make builds more modular and allow a wider variety of devices to run on the same binary image.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Discussion of the ongoing effort within the Android project to make the build and binaries more modular with the goal to make device bringup easier and faster.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Rom Lemarchand</b></p><div/><p>Rom Lemarchand,<br />
Staff Engineer, Android Systems.</p>
<p>Rom has been working in the consumer electronics industry for over 10 years. Originally working on various proprietary operating systems, he switched to Android about 6 years ago.</p>
<p>Rom joined the Android team at Google in 2012 and since then has been working on a variety of subsystems (graphics, memory…) and devices (Nexus 9, Android One…).</p>
<p>Rom holds a degree in Computer Science from Universite de Rouen and a Masters in Software Engineering from Carnegie Mellon University.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33812015-07-15T00:07:02-07:002015-07-15T00:07:02-07:00Open-Channel SSDs2015-07-15T07:07:02Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Support for Open-Channel SSDs - Ongoing progress</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Support for Open-Channel SSDs is an open-source initiative to push management of SSDs into the host. The initiative enables application-driver storage, where applications directly integrate with the underlying storage and its characteristics. Leading to improvements in storage utilization, predictability and latency. We’ll outline the current roadmap, missing pieces before going upstream, and upcoming vendors and applications supporting the framework.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Matias Bjørling</b></p><div/><p>Matias Bjørling obtained a Ph.D. in operating systems and characteritics of solid state drives. He is a member of <span class="caps">CNEX</span> Labs technical staff. Prior to that, he worked on performance characterization of flash-based SSDs, operating- and database systems research and the multi-queue block layer. Matias is a contributor to the Linux Kernel and is the creator of the LightNVM framework for Open-Channel SSDs.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33692015-07-14T16:08:28-07:002015-08-24T10:37:18-07:00Reducing the ARM Linux kernel size without losing your mind2015-07-14T23:08:28Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Ways to reduce the Linux kernel size and the problems that they face.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Automatically reducing the Linux kernel size may be achieved in<br />
different ways. Using <span class="caps">LTO</span> (Link Time Optimization) is one such way with many advantages, but it also has major issues. A simpler alternative is linker section garbage collection. However, it turns out that even “simpler” solutions have their share of unsuspected pitfalls, especially on <span class="caps">ARM</span>. Those pitfalls and proposed solutions are the subjects of this discussion.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Nicolas Pitre</b></p><div/><p>Nicolas Pitre has been a Linux developer on the <span class="caps">ARM</span> architecture since <br />
1995. He started maintaining the Linux port for the StrongARM SA1100 in 1997 and has been an active member of the <span class="caps">ARM</span> Linux community since then. In the last 15 years, he worked closely with a few silicon vendors to add Linux support for various <span class="caps">ARM</span> based products while adding some core features to the <span class="caps">ARM</span> kernel such as <span class="caps">EABI</span> and highmem. His latest contributions revolved around <br />
power management for ARM’s big.<span class="caps">LITTLE</span>. Code efficiency and binary size has always been a pet peeve for Nicolas which also prompted him to contribute improvements to gcc on a few occasions.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33632015-07-14T16:00:53-07:002015-07-21T05:58:25-07:00Securing the Entire Boot Chain 2015-07-14T23:00:53Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion around a trusted boot chain and how to assure your system is running the software that you intend.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Secure Boot is a single component of designing a boot process that allows users to feel secure that their systems have not been tampered with. Several other technologies (TPMs, <span class="caps">TXT</span>, disk encryption, remote attestation and associations between them) exist, but often require extensive manual configuration and fail to provide a seamless user experience. What would a fully secure boot chain look like, and what should we be doing to get there?"</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Matthew Garrett</b></p><div/><p>Matthew Garrett is a security developer at CoreOS, developing technologies to improve the security of containers and the systems that run them. He has a background in firmware integration, power management and fruitfly genetics and so has atypical ideas about system complexity and the ease of reverse engineering. A board member at the Free Software Foundation and a a member of the Linux Foundation Technical Advisory Board, he has strong feelings on high-quality Free Software.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33572015-07-09T18:44:35-07:002015-09-03T17:43:38-07:00Designing Plugin Systems for Container Runtimes2015-07-10T01:44:35Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Containers are a spectrum of options and every system wants to configure things differently; to make this configurable runtimes like docker and rkt are using plugins.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Network plugins for application runtimes like rkt and docker are used to customize the network namespace before a process is executed. I will cover the basic use cases of container network setup, the considerations of creating and configuring a network namespace, and plumbing issues encountered and worked around in the design and implementation. Today this discussion revolves around two options: github.com/docker/libnetwork and github.com/appc/cni. Overtime the two approaches should converge.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Brandon Philips</b></p><div/><p>Brandon Philips is helping to build modern Linux server infrastructure at CoreOS as <span class="caps">CTO</span>. Prior to CoreOS, he worked at Rackspace hacking on cloud monitoring and was a Linux kernel developer at <span class="caps">SUSE</span>. As a graduate of Oregon State’s Open Source Lab he is passionate about open source technologies.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33512015-07-09T17:42:23-07:002015-07-19T09:16:52-07:00Running Docker inside VZ containers2015-07-10T00:42:23Z<div><dl><dt><b>One Line Summary:</b></dt><dd>The session summarizes our experience in making Docker working inside VZ containers.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The session focuses on running Docker inside VZ containers:</p>
<ul>
<li>what makes our use case different from others — we have to control how Docker accesses to the resources of host system</li>
<li>devicemapper graphdriver and proxy graphdriver</li>
<li>how to pass a mount point made on host system to container namespace</li>
</ul></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Maxim Patlasov</b></p><div/><p>Maxim Patlasov currently works as OpenVZ kernel team developer, specializing in filesystem, storage and cluster technologies. Previously, he had been working on network part of Lustre filesystem. He has more than seven years of Linux experience and holds a M.Sc. degree in Applied Mathematics from the Moscow State University.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33452015-07-09T15:38:04-07:002015-07-19T09:16:37-07:00Containers in the upstream kernel (as compared to VZ kernel)2015-07-09T22:38:04Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Comparison of containers features between OpenVZ and upstream kerne</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Historically, we have developed our own OpenVZ/Virtuozzo kernels as fork of the upstream (and later <span class="caps">RHEL</span>) kernels, with massive patchsets and lots of added functionality. For the last 10 years, though, we work hard to bring these features to the upstream kernel, and it pays both to us (less code to port) as well as to others (say, <span class="caps">LXC</span> and Docker can now use vanilla kernel). While upstream kernel is good enough, it’s still not on par with the OpenVZ one.</p>
<p>We have just finished rebasing OpenVZ kernel from RHEL6 (2.6.32-based) to RHEL7 (3.10-based) kernel. This session summarizes our experience and findings,<br />
including answers to what is still missing from upstream kernel. The session also outlines our plans<br />
to contribute to upstream.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Kir Kolyshkin</b></p><div/><p>Developing free software on Linux since 1998. Working on Linux Containers since 2002. Leading the OpenVZ project since 2005.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33392015-07-09T15:30:44-07:002015-07-19T09:16:20-07:00Checkpoint/restore status update (CRIU)2015-07-09T22:30:44Z<div><dl><dt><b>One Line Summary:</b></dt><dd>CRIU status update, incl. integration (Docker/LXC/LXD/VZ7)</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Checkpoint/restore is a technology that was OpenVZ/Virtuozzo users enjoyed for a long time, but it has made its way to upstream kernel only recently, thanks to <span class="caps">CRIU</span> project. In fact it’s so popular that there’s a whole Checkpoint/Restore microconference. This session, though, tries to summarize the current status of <span class="caps">CRIU</span>, including integrating it into various container management tools such as Docker, <span class="caps">LXC</span>, Virtuozzo 7 etc.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Kir Kolyshkin</b></p><div/><p>Developing free software on Linux since 1998. Working on Linux Containers since 2002. Leading the OpenVZ project since 2005.</p></li><li><p><b>Pavel Emelyanov</b></p><div/><p>Pavel Emelyanov is a core Virtuozzo and OpenVZ kernel developer working in the project for the last six years. He’s currently the kernel team leader at Parallels and manages the development of all the kernel features for OpenVZ and Parallels Virtuozzo Containers. Pavel is also a prolific mainstream kernel contributor. He holds a PhD degree in Applied Mathematics from the Moscow Institute of Physics and Technology.</p></li><li><p><b>Andrey Vagin</b></p><div/><p>Developer in the <span class="caps">CRIU</span> and OpenVZ projects.</p>
<p>Andrew Vagin is interested in Container Virtualization (<span class="caps">LXC</span>, OpenVZ). He started to write autotest for OpenVZ in 2006, when he was a student at the Moscow Institute of Physics and Technology (<span class="caps">MIPT</span>). Now Andrew works in OpenVZ kernel team. In addition he is an active developer in the <span class="caps">CRIU</span> (Checkpoint/Restore in Userspace) project.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33332015-07-09T15:24:14-07:002015-07-19T09:17:14-07:00cgroups kernel memory controller2015-07-09T22:24:14Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Kernel memory accounting (kmemcg): status, why important, why and how to use</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Containers need resource management and limiting. One of such resources that was overlooked until before recently is kernel memory.</p>
<p>On almost all operations involving interactions with the kernel, it implicitly allocates some memory, that needs to be accounted and controlled. This is something that was present in OpenVZ kernel for more than 10 years (kmemsize user beancounter), but was only recently merged upstream by Virtuozzo engineers, in the form of kmem cgroup controller.</p>
<p>The topic discusses the current status of kmem controller (what is done and what is still missing), why it’s important to use it (including what sorts of attacks can be prevented) and how to use it properly and effectively from the userspace container management tools such as Docker, <span class="caps">LXC</span>, CoreOS etc.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Kir Kolyshkin</b></p><div/><p>Developing free software on Linux since 1998. Working on Linux Containers since 2002. Leading the OpenVZ project since 2005.</p></li><li><p><b>Pavel Emelyanov</b></p><div/><p>Pavel Emelyanov is a core Virtuozzo and OpenVZ kernel developer working in the project for the last six years. He’s currently the kernel team leader at Parallels and manages the development of all the kernel features for OpenVZ and Parallels Virtuozzo Containers. Pavel is also a prolific mainstream kernel contributor. He holds a PhD degree in Applied Mathematics from the Moscow Institute of Physics and Technology.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33212015-07-09T13:49:17-07:002015-08-12T04:43:34-07:00BOF: Device Mapper for Shingle Magnetic Recording (SMR)2015-07-09T20:49:17Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Approach and progress in developing a device mapper for SMR</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Abstract<br />
Any problem in computer science can be solved with another layer of indirection. Shingle magnetic recording is no different – the only “difficulty” is to determine where to add the additional layer of indirection/abstraction to enable maximum flexibility and efficiency. Let’s go over the various SW/FW paradigms that attempt to abstract away <span class="caps">SMR</span> behavior (e.g. user space library, device mapper, <span class="caps">SMR</span> aware file system, enlightened application). Along the way, we will also explore what deficiencies are holding back <span class="caps">SMR</span> adoption in (e.g. <span class="caps">ATA</span> sense data reporting) the data center.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>jim malina</b></p><div/><p>Jim Malina is a Technologist of System Architecture at Western Digital. I’ve made a career finding simple solutions for complex problems. Currently evangelizing “storage stack” solutions for next generation of recording media. Previously at Texas Instruments Digital Light Processing, developing system solutions for light modulation devices in display and communication technologies. Began my career at Seagate’s R&D facility in Los Angeles.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33152015-07-09T08:06:08-07:002015-09-03T17:43:49-07:00Open Container Specifications2015-07-09T15:06:08Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Summary and future direction of the open container specifications for application focused containers</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Application containers are changing some of the fundamentals of how Linux is used in the server environment. An application container differs from a system container in that the root filesystem for the container is designed to launch a single application instead of a full init.</p>
<p>The open container specifications are an attempt to write-down an interoperable set of specifications for use by different application container runtimes; on Linux this includes Docker, kurma, and rkt. The project involves engineers from a number of different projects and the current status is work-in-progress but there should be some initial releases and significant progress in time for <span class="caps">LPC</span>.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Brandon Philips</b></p><div/><p>Brandon Philips is helping to build modern Linux server infrastructure at CoreOS as <span class="caps">CTO</span>. Prior to CoreOS, he worked at Rackspace hacking on cloud monitoring and was a Linux kernel developer at <span class="caps">SUSE</span>. As a graduate of Oregon State’s Open Source Lab he is passionate about open source technologies.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33092015-07-08T16:04:28-07:002015-08-12T04:44:04-07:00BOF: Kernel networking and TCP's (ir)relevance in the latest brave new world2015-07-08T23:04:28Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Let's talk about the latest trend in transport protocol development which is to expressly avoid relying on the kernel</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Defining transport protocols running on userspace over <span class="caps">UDP</span> is an emerging trend. <span class="caps">QUIC</span> is starting to be deployed, <span class="caps">SPUD</span> is being discussed in <span class="caps">IETF</span>, there are efforts to implement <span class="caps">TCP</span> and <span class="caps">SCTP</span> over <span class="caps">UDP</span> in userspace. A key motivation for this trend is to avoid the kernel like the plague. As stated in the <span class="caps">QUIC</span> wiki: “As a userspace transport atop <span class="caps">UDP</span>, <span class="caps">QUIC</span> allows innovations which have proven difficult to deploy with existing protocols as they are hampered by legacy clients and middleboxes, or by prolonged Operating System development and deployment cycles.”</p>
<p>In this <span class="caps">BOF</span>, we will raise at least two questions</p>
<p>1) Is there anything we can do from the kernel perspective to address the “prolonged development and development cycles” issue?<br />
2) Other than continuing to provide functional <span class="caps">UDP</span> sockets, is there anything we should do in the kernel to support transport protocols in userspace model?</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Tom Herbert</b></p><div/><p>Software engineer at Facebook and Linux netdev contributor.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/33032015-07-07T04:38:40-07:002015-07-07T04:38:40-07:00Switchdev driver for ASIC2015-07-07T11:38:40Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A discussion about the implementation of a switchdev driver for advanced ASIC</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The switch device driver model (switchdev) enables the offloading of the forwarding plane from kernel to the hardware.<br />
In this talk we will present a basic switch driver with provisions for L2 and L3 offloads and discuss key issues in the implementation, such <span class="caps">FDB</span> aging and exceeding hardware resources.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Elad Raz</b></p><div/><p>Experienced entrepreneur and business leader who is well respected for his technical expertise and top-notch project management skills. Specializes in the fields of low-level development, information security, networking and file-system drivers, with a unique system-wide knowledge of many low-level platforms including Intel x86_64 platforms, <span class="caps">ARM</span>-based platforms, Microsoft Windows internals and Linux internals.</p></li><li><p><b>Ido Schimmel</b></p><div/><p>Software Engineer @ Mellanox</p></li><li><p><b>matty kadosh</b></p><div/><p>matty kadosh is a software architect at mellanox where he works on various networking components focusing on Linux and switch software.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32972015-07-06T13:41:42-07:002015-07-07T15:14:07-07:00Overlay Encryption2015-07-06T20:41:42Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Ensuring that new network architectures have built-in security</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Overlay based network virtualization provides a nice way to enable ubiquitous encryption in the data center since it avoids many of the pain points that have historically prevented its deployment. However, while the pieces are mostly there or are coming in the near future, it’s not yet something that is trivial to implement.</p>
<p>This is a follow on to last year’s talk, specifically focusing on what can be done with new protocols and hardware that is starting to emerge.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Jesse Gross</b></p><div/><p>Contributor to Linux kernel networking and Open vSwitch as well as coauthor of various protocols used for network virtualization.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32912015-07-02T21:11:28-07:002015-08-21T21:41:57-07:00Unfair Queued Spinlocks and Transactional Locks2015-07-03T04:11:28Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A performance review on unfair queued spinlocks for virtual guests and transactional locks based on the Intel's TSX technology.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The qspinlock patch that went into the 4.2 kernel had support for PV qspinlock for paravirtual guests and simple byte unfair lock for other VM guests. On an over-committed guest, the heavy vmexit and vmenter overhead of pvqspinlock can slow down system performance rather significantly. The unfair byte lock performs well on small guest, but it doesn’t scale well when the number of vCPU increases. The unfair qspinlock patch can scale almost as well as the fair qspinlock patch without the performance overhead of the pvqspinlock.</p>
<p>On Intel Haswell-EX and newer CPUs, the <span class="caps">TSX</span> technology can enable a new class of transactional locks that can be useful in situation where fine-grained locking is hard. This session will talk about a proposed implementation for a new set of locking APIs that are direct replacement of the spinlock APIs to enable the use of transactional locking for selected locks in the kernel.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Wai Man Long</b></p><div/><p>Waiman Long is an experienced software engineer working in the Linux Performance and Scalability team of HP Enterprise Server Group. He is responsible for investigating and resolving performance and scalability issues of the Linux kernel on large multi-node <span class="caps">NUMA</span> systems. He is also a major contributor in revamping the Linux kernel synchronization primitives in recent years.</p>
<p>He has multiple years of working experience on Linux, HP-UX and Tru64 <span class="caps">UNIX</span> in both the kernel and the C runtime library. He is also an expert in the field of software internationalization.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32852015-07-01T15:23:06-07:002015-08-17T07:08:11-07:00BOF: switchdev offloading 2015-07-01T22:23:06Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A look on the common dilemma for offloading Linux network component to HW switches and e-switches using switchdev </dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The switch device driver model (switchdev) enables offloading of the forwarding plane from the kernel to the hardware tables, where the HW forwarding plane is viewed as a fast path cache of the Linux networking tables.<br />
In this session, we discuss some issues that a vendor may face when implementing a switchdev offloading driver:<br />
Offload policy options:<br />
- Full synchronization (offload all)<br />
- Traffic based (cache)<br />
- <span class="caps">LRU</span> based<br />
- Other?<br />
How/should the application control the offload policy<br />
Out Of Resource management:<br />
What should be the behavior in case the HW resources exhausted?<br />
- Ignore HW error and leave kernel and HW out of synched ? in that case, how do we synch the forwarding plane back again when HW resource become available ?<br />
- Force full synchronization of kernel and HW tables i.e. delete the failed entry’s form the kernel as well?<br />
- Provide some priority mechanism to replace “less important” entries with more important ones? <br />
- How to implement a per table priority policy?</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>matty kadosh</b></p><div/><p>software architect at Mellanox works on various networking components focusing on HW switches <span class="caps">SDK</span> and driver ,Linux, network protocols ,and <span class="caps">SDN</span>.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32792015-07-01T12:00:35-07:002015-07-01T12:52:31-07:00Live sources and sinks in KMS2015-07-01T19:00:35Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Support for live/stream-based sources and sinks within KMS</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p><span class="caps">DRM</span> and <span class="caps">KMS</span> have always dealt very specifically in terms of buffers and frames, rather than connecting arbitrary streams together. Dealing in streams has always been the domain of userspace, with frameworks such as GStreamer and OpenMAX IL.</p>
<p>Unfortunately some hardware can only work in a stream-based manner, such as media decode engines (or cameras) to display engines, and also ‘writeback’ from display engine outputs back to media encode engines.</p>
<p>This session will look at the work required to support live sources and sinks within <span class="caps">KMS</span>.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Laurent Pinchart</b></p><div/><p>Laurent Pinchart has been a Linux kernel developer since 2001. He has written media-related Linux drivers for consumer and embedded devices, and is one of the Video4Linux core developers. Laurent is the founder and owner of Ideas on board, a company specialized in embedded Linux design and development. He gave Linux media-related talks at the <span class="caps">FOSDEM</span>, <span class="caps">LPC</span> and <span class="caps">ELC</span>. He currently works with the Renesas Linux kernel team where he develops embedded video-related <span class="caps">DRM</span>/<span class="caps">KMS</span> and V4L2 drivers.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32732015-07-01T11:58:12-07:002015-09-01T03:43:24-07:00ION Feature Upstreaming: Issues that remain2015-07-01T18:58:12Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Working towards a common generic display-capable buffer allocation API</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>One significant pain point for vendors is the deeply platform-specific knowledge required to allocate buffers which will be suitable for any combination of media decode engines, display engines, GPUs, and similar.</p>
<p>Various ideas have been discussed for <span class="caps">ION</span> allocator replacement, and some RFC’ed – lets revisit them and see if we’re on the right path, or is there a path-correction required.</p>
<p>Also, discuss the other <span class="caps">ION</span> issues that still remain that must be handled for an upstream-worthy solution.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Sumit Semwal</b></p><div/><p>Sumit leads a small team inside Linaro Mobile Group to work on Android Kernel Upstreaming. In addition,<br />
Sumit is the maintainer of dma-buf buffer sharing framework in the Linux kernel.</p>
<p>Earlier, Sumit worked with Texas Instruments in their <span class="caps">OMAP</span> division, in various engineering and techno-managerial capacities.</p>
<p>Sumit holds a Masters in Computer Applications with specialization in Software Engineering, from IP University of Delhi.</p>
<p>Sumit loves the exposure to multiple cultures that working with the global Linux community gets him. oh, and of course, the travel :).</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32672015-07-01T11:56:06-07:002015-08-16T05:15:08-07:00Gap between Atomic Modesetting / KMS and ADF2015-07-01T18:56:06Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Looking at the KMS atomic modesetting API and examining the feature gap to Android's ADF</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>With the 4.2 kernel, the <span class="caps">DRM</span>/<span class="caps">KMS</span> atomic modesetting <span class="caps">API</span> is finally enabled by default, allowing userspace to submit whole configurations across a number of display elements, rather than the previous incremental approach. This session will look at the remaining feature and implementation work to be done on <span class="caps">AMS</span>, including the feature gap to Android’s Atomic Display Framework (<span class="caps">ADF</span>), and how to enable easy porting of Android drivers and userspace.</p>
<p>For a design over of atomic mode setting please read <a href="https://lwn.net/Articles/653071/" rel="nofollow">https://lwn.net/Articles/653071/</a> and <a href="https://lwn.net/Articles/653466/" rel="nofollow">https://lwn.net/Articles/653466/</a></p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Daniel Vetter</b></p><div/><p>Daniel Vetter is the maintainer of the Intel drm/i915 graphics driver since a few years. Besides the usual maintainer duties and feature work he spends a sizeable amount of time in trying to rectify past mistakes in the drm subsystem. Of which there are a lot, and a lot of these mistakes are in the form of bad userspace <span class="caps">API</span> which can’t be removed anytime soon. He is working at the Intel Open Source Technology Center.</p></li><li><p><b>Rob Clark</b></p><div/><p>Rob works in the (too small) intersection of open source graphics and <span class="caps">ARM</span>, writing open src gallium driver for snapdragon/adreno 2xx and 3xx GPUs, and <span class="caps">EXA</span> 2d support for z180. And currently writing his 3rd drm/kms driver.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32612015-06-30T21:26:54-07:002015-08-22T14:06:23-07:00Beyond traditional compilation 2015-07-01T04:26:54Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Beyond traditional compilation</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Beyond traditional compilation <br />
or why the Linux community should stop the single compiler monopoly.</p>
<p>Compilers are not only for code generation any more.<br />
Modern compilers provide a wide range of tools that perform static and dynamic program analysis, control- and data-flow-guided fuzzing, and security hardening. We’ll talk about all of these new compiler features implemented in Clang/<span class="caps">LLVM</span> and show how they are going to help Linux, both the kernel and the user-space.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Kostya Serebryany</b></p><div/><p>Konstantin (Kostya) Serebryany is a Software Engineer at Google. During last ~ 8 years his team develops and deploys dynamic testing tools, such as AddressSanitizer and ThreadSanitizer. Prior to joining Google, Kostya spent 7 working on compiler optimizations at Sun and Intel. <br />
<a href="http://research.google.com/pubs/KonstantinSerebryany.html" rel="nofollow">http://research.google.com/pubs/KonstantinSerebryany.html</a></p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32552015-06-30T20:34:34-07:002015-07-02T04:50:47-07:00network virtualization with BPF2015-07-01T03:34:34Z<div><dl><dt><b>One Line Summary:</b></dt><dd>how to create virtual networks by gluing kernel components with BPF</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The linux kernel already has software bridges, routers, tunnels and others building blocks that can be put together to construct fast and scalable virtual networks for cloud and container environments. This talk describes how one can use a little bit of TC+BPF glue code to connect them into network topologies and beyond.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Alexei Starovoitov</b></p><div/><p>Engineer @ PLUMgrid</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32492015-06-30T18:13:39-07:002015-08-22T08:42:20-07:00BPF in LLVM and kernel2015-07-01T01:13:39Z<div><dl><dt><b>One Line Summary:</b></dt><dd>status of eBPF development in LLVM, clang, kernel and other tools</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The extended <span class="caps">BPF</span> (eBPF) architecture brought the power of on-the-fly generated and optimized code by <span class="caps">LLVM</span> into the linux kernel. It’s currently used by tracing and networking subsystems. This talk describes compiler related parts of eBPF: code generation in the backend, object file & loaders, in-kernel <span class="caps">JIT</span>, <span class="caps">LLVM</span> <span class="caps">JIT</span>, clang rewriter, c+python hybrid, other front-ends.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Alexei Starovoitov</b></p><div/><p>Engineer @ PLUMgrid</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32432015-06-30T16:23:26-07:002015-08-22T08:43:08-07:00Limitations of various linkers in clang based toolchains2015-06-30T23:23:26Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Limitations of various linkers in clang based toolchains</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Experiences using various linkers (ld.bfd, ld.gold, MCLinker, and llvm-link) for building the Dronecode PX4 Flight stack when porting to Linux.</p>
<p>The PX4 code makes use of some features of ld.bfd that are difficult to work around using other linkers. This session will be a discussion of the various issues discovered and options to work around the issues.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Mark Charlebois</b></p><div/><p>Mark Charlebois is a Director of Open Source Software Strategy at Qualcomm Innovation Center, Inc. (QuIC), a subsidiary of Qualcomm, Inc. He has been developing software for Unix, Linux and embedded systems for more than 20 years. Mark joined Qualcomm in 1999, where he initially worked on the GlobalStar satellite system. He subsequently led one of the core subsystem teams for MediaFLO, defining core parts of the system information protocol. He has worked for several years in various R&D organizations at Qualcomm and created the initial Penelope extension to Thunderbird that formed the basis for Open Source Eudora. While in R&D, Mark was an open source evangelist who helped drive the company’s open source strategy. He has moved to QuIC to take a more active role in defining its open source strategy and community engagements. Mark currently serves on the Board of the Linux Foundation. Recently he has been working on compiling the Linux kernel with Clang and helping to maintain and upstream patches to support that goal via the LLVMLinux project. Prior to Qualcomm, Mark worked at <span class="caps">LMI</span> in greater Vancouver, designing laser scanners. He also worked for <span class="caps">CGI</span> in Toronto as a real-time systems consultant using various Unix systems. Mark holds a bachelor’s degree in Systems Design Engineering from the University of Waterloo, and a master’s degree in Engineering Science from Simon Fraser University.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32372015-06-30T13:24:01-07:002015-08-22T08:33:52-07:00Using clang static analyzer with the Linux kernel code2015-06-30T20:24:01Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Demo of how to use the clang static analyzer with the Linux kernel code.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Static analyzers can mechanically find issues with code that would otherwise take humans a lot longer to figure out. By using a static analyzer which is built using the same technology as the compiler gives the analyzer the ability to find potentially complex cross functional bugs and other issues. This will be a demo of how developers can use the clang static analyzer to help find potential issues with kernel code.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Behan Webster</b></p><div/><p>Behan Webster has spent two decades in diverse tech industries such as telecom, datacom, optical, and automotive writing code for a range of hardware from the very small to the very large. He has Linux experience spans kernel programming, Embedded Linux, and board bring-up. Currently Behan is the lead consultant at Converse in Code Inc, an embedded Linux engineer and project lead working on the LLVMLinux project as well as being a Trainer for The Linux Foundation.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32312015-06-30T13:23:55-07:002015-08-22T08:33:19-07:00Using clang static analyzer with the Linux kernel code2015-06-30T20:23:55Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Walkthrough of how to use the clang static analyzer with the Linux kernel code.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Static analyzers can mechanically find issues with code that would otherwise take humans a lot longer to figure out. By using a static analyzer which is built using the same technology as the compiler gives the analyzer the ability to find potentially complex cross functional bugs and other issues. This session will allow people to build the kernel with clang and use clang static analyzer to help find potential issues with kernel code.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Behan Webster</b></p><div/><p>Behan Webster has spent two decades in diverse tech industries such as telecom, datacom, optical, and automotive writing code for a range of hardware from the very small to the very large. He has Linux experience spans kernel programming, Embedded Linux, and board bring-up. Currently Behan is the lead consultant at Converse in Code Inc, an embedded Linux engineer and project lead working on the LLVMLinux project as well as being a Trainer for The Linux Foundation.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32252015-06-30T13:21:39-07:002015-08-22T10:04:18-07:00Done ? Final spurt ? - Compiling the Linux Kernel with clang on x862015-06-30T20:21:39Z<div><dl><dt><b>One Line Summary:</b></dt><dd>The LLVMLinux project upstreamed a lot of patches to the Linux Kernel (and LLVM) - are we done and what is left ?</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Going forward we need to get the last missing pieces into mainline so more developers can actually use it out of the box.<br />
Unfortunately we’re not done yet for the Linux kernel and a lot of small issues remain in the drivers. <br />
This talk summarizes the issues and aims answer the question of the priorities for the remaining issues based on the feedback of the audience.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Jan-Simon Möller</b></p><div/><p>Jan-Simon Möller is a consultant and trainer for the Linux Foundations Training Program. He holds a Dipl.-Ing. degree in electrical engineering with strong background in Embedded Linux ranging from Realtime to SELinux.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32192015-06-29T15:01:42-07:002015-07-01T13:47:38-07:00SMR in the kernel - Work remaining2015-06-29T22:01:42Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A litany of work remaining for enabling SMR in the Kernel.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>We’ll explore a quick summary of what we (Seagate) have done for the adoption of <span class="caps">SMR</span> drives, another summary of the proposed changes outstanding that we have scoped, and a list of designs that need some initial thought and scoping.</p>
<p>Addressable topics include:<br />
HBAs – vendor specific</p>
<p>libata/libsas/scsi – mostly done, thanks to Dr. Reinecke and Seagate engineers</p>
<p>drivers – sd (and st?) —Difference between tape, flash, <span class="caps">SMR</span></p>
<p>blockdev changes – again, Dr. Reinecke and Seagate engineers</p>
<p>IO elevator for IO order</p>
<p>FS design (based off ext4): major algorithm changes<br />
block allocator<br />
garbage collector – compactor/defragmentor<br />
b+trees (dynamic inodes, dynamic group sizes, group allocation/purpose updates)<br />
Journaling</p>
<p>DM shims – {HM,Conv} to {HM,Conv}</p>
<p>mdraid/<span class="caps">LVM</span> – Buttressing, Overlapping, Interleaving, and Parity: aggregation of zones to FS.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Adrian Palmer</b></p><div/><p>Adrian Palmer holds a MS of Electrical Engineering and a BS of Computer Engineering and Math, all from the University of Wyoming. He has spent over 10 years in Systems Administration, and has hobbyist experience in Linux and <span class="caps">UNIX</span> systems, creating router/gateway and <span class="caps">NAS</span> systems. In 1999, he was the first public high school student to earn an <span class="caps">MCSE</span>. Adrian currently works at Seagate and leads the <span class="caps">SMRFFS</span> project, contributing to the open source community.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32132015-06-29T11:03:19-07:002015-06-29T15:07:27-07:00Identifier locator addressing2015-06-29T18:03:19Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A method to do network virtualization without encapsulation by leveraging the insanely large address space of IPv6.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Identifier Locator Addressing (or <span class="caps">ILA</span>) is a mechanism to implement tunnels or overlay networks that does not require encapsulation. This is done by splitting an IPv6 address into a locator part and an identifier part. The identifier (who) indicates a virtual node in the network, and the locator (where) indicates the location of the node.</p>
<p><span class="caps">ILA</span> is intended to be a key component in virtualizing everything in a data center. This can be thought of as giving every task or container its own IP address and allow seamless migration for that.</p>
<p>In this talk we will discuss the implementation of <span class="caps">ILA</span> in the Linux kernel. The key requirements are:<br />
– Near zero overhead in the data path. <br />
– MInimal invasiveness in the stack<br />
– Scalable interface managing mappings<br />
– Protocol correctness</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Tom Herbert</b></p><div/><p>Software engineer at Facebook and Linux netdev contributor.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32072015-06-29T09:00:00-07:002015-06-29T15:08:34-07:00IPv6 and 6LoWPAN only networking for sensor devices2015-06-29T16:00:00Z<div><dl><dt><b>One Line Summary:</b></dt><dd>IPv6 and 6LoWPAN only networking for sensor devices</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Modern IoT sensor devices will operate utilizing 6LoWPAN over 802.15.4 and Bluetooth Low Energy transports. They only require IPv6 and are limited in resources. The topic is on how to create a Linux networking configuration that suits these small devices.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Marcel Holtmann</b></p><div/><p>Marcel Holtmann is working at Intel’s Open Source Technology Center. He is the maintainer of the BlueZ open source Bluetooth stack and has been working on Bluetooth technology since 2001.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/32012015-06-28T14:23:28-07:002015-06-29T06:13:08-07:00IPv6: Routing Cache Removal and Going Forward2015-06-28T21:23:28Z<div><dl><dt><b>One Line Summary:</b></dt><dd>The recent IPv6 routing cache removal and other IPv6 improvements we are looking at in Facebook.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Facebook has a huge IPv6 deployment. Our internal DC traffic is mostly IPv6 and we also support end-user IPv6 traffic. This session will talk about the scalability issue we had and the recent routing cache removal from IPv6. It will also talk about other opportunities that we are looking at.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Martin KaFai LAU</b></p><div/><p>Software Engineer at Facebook</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31952015-06-28T11:09:42-07:002015-06-28T11:09:42-07:00Geneve Openflow API2015-06-28T18:09:42Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Geneve Openflow API</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Openvswitch Geneve openflow <span class="caps">API</span> and tunnel metadata format.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Madhusudan Challa</b></p><div/><p>Linux networking, performance.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31892015-06-27T17:29:10-07:002015-06-29T06:12:55-07:00VRF's in linux : reusing infrastructure2015-06-28T00:29:10Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Proposal to implment VRF lite using existing linux constructs</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>In the context of internet scale routing a requirement that always comes up is the need to partition the available routing tables into disjoint routing planes. A specific use case is the multi-tenancy problem where each tenant has their own unique routing tables and in the very least need different default gateways.</p>
<p>This is an attempt to build the ability to create virtual router domains aka VRF’s (<span class="caps">VRF</span>-lite to be specific) in the linux packet forwarding stack. The main observation is that through the use of rules and socket binding to interfaces, all the facilities that we are there in the infrastructure. What is missing is a handle that identifies traffic that is associated with a <span class="caps">VRF</span> that preserves the notions of <span class="caps">ECMP</span>, and general routing principles.</p>
<p>The driver is a cross between functionality that the <span class="caps">IPVLAN</span> driver and the VXLan drivers provide where a device is created and packets into/out of the routing domain are shuttled through this device. The device is then used as a handle to identify the applicable rules. The <span class="caps">VRF</span> device is thus the layer3 equivalent of a vlan device.</p>
<p>High Level aspects of proposal :<br />
Simple overlay driver (minimal changes to current stack)<br />
Uses the existing fib tables and fib rules infrastructure<br />
Modeled closely after the ipvlan driver<br />
Uses current <span class="caps">API</span> and infrastructure.<br />
Applications can use SO_BINDTODEVICE or cmsg device indentifiers to pick <span class="caps">VRF</span> (ping, traceroute just work)<br />
Standard IP Rules work, and since they are aggregated against the device, scale is manageable<br />
Completely orthogonal to Namespaces and only provides separation in the routing plane (and <span class="caps">ARP</span>)<br />
Debugging is built-in as tcpdump and counters on the <span class="caps">VRF</span> device works as is.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Shrijeet Mukherjee</b></p><div/><p>VP Software, Cumulus Networks</p></li><li><p><b>David Ahern</b></p><div/><p><span class="caps">MTS</span> at Cumulus Networks working on the Linux kernel.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31832015-06-26T02:44:53-07:002015-06-29T06:11:23-07:00eBPF and tc classifier/actions2015-06-26T09:44:53Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Overview and next steps on eBPF tc integration.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>This talk/discussion provides an overview and next steps on eBPF integration into the network scheduling layer of the Linux kernel and iproute2’s tc user space tool.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Daniel Borkmann</b></p><div/><p>Contributor to the Linux kernel and various other Open Source projects.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31772015-06-25T11:31:09-07:002015-08-21T00:35:49-07:00Research topics in fastboot2015-06-25T18:31:09Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Examine low hanging fruit for sub-second booting of systems</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>There are many opportunities to improve the boot time of Linux systems. This talk will examine some of the more promising<br />
1. Deferring initcalls. There is already a mechanism to run the initcalls for a module at module installation. This talk will examine how to extend that mechanism for built-in kernel subsystems to defer running the initcall off the fastpath.<br />
2. Examine ways to start drivers in u-boot on multicore CPUs. The driver could run on a core in <span class="caps">DSP</span> mode until the kernel was ready to take over its responsibilities.<br />
3. Parallelize <span class="caps">CPU</span> and I/O operations. Examples include decompressing the kernel in parallel with loading it, starting the mount of the rootfs early in kernel bringup, and preloading of applications.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>John Mehaffey</b></p><div/><p>John Mehaffey has been working on embedded Linux since 1989. He is currently a Linux architect at Mentor Graphics working on fastboot.</p>
<p>John designed real-time computers at Hewlett-Packard, and has a deep understanding of the boot and early system bringup process. John presented most recently at <span class="caps">ELC</span> about fastboot.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31712015-06-25T10:47:20-07:002015-08-12T04:45:06-07:00BOF: Virtual Routing Framework (VRF)2015-06-25T17:47:20Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion of VRF architecture and implementation issues</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>There are multiple possible ways to support multiple Virtual Routers on a single system. The changes impact the networking stack on many levels. Recently discussion between multiple companies and users has centered on how to make a scalable solution. Any result impacts Linux userspace infrastructure and kernel.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Stephen Hemminger</b></p><div/><p>Stephen has been involved with networking and <span class="caps">SMP</span> for many years.<br />
and has been active in the kernel networking community for 7 years.<br />
In addition to working for Vyatta, in developing their open source routing<br />
platform he has contributed the “netem” network emulator, several<br />
network drivers and is the maintainer of the bridging and iproute utilities.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31652015-06-25T08:50:58-07:002015-08-25T02:48:45-07:00P.Haul: live migrating with CRIU2015-06-25T15:50:58Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A quick talk about p.haul project that live migrates containers using CRIU</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Doing live migration is not just dump + copy + resume, to decrease the freeze time migration should include some steps that copy data between nodes without freezing the tasks.</p>
<p><span class="caps">CRIU</span> provides all necessary bits and pieces to do it, but doesn’t allow to easily do all of them with one simple call. The P.Haul project is all about it.</p>
<p>In this talk I’d like to describe the current P.Haul architecture and discuss what we plan to do with it to suit all the interested parties.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Pavel Emelyanov</b></p><div/><p>Pavel Emelyanov is a core Virtuozzo and OpenVZ kernel developer working in the project for the last six years. He’s currently the kernel team leader at Parallels and manages the development of all the kernel features for OpenVZ and Parallels Virtuozzo Containers. Pavel is also a prolific mainstream kernel contributor. He holds a PhD degree in Applied Mathematics from the Moscow Institute of Physics and Technology.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31592015-06-24T14:24:44-07:002015-08-29T00:05:37-07:00Scheduler-driven CPU frequency selection2015-06-24T21:24:44Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discuss current design and gather feedback on a cpu capacity-based frequency scaling policy.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Discuss design of a scheduler-driven cpufreq policy based on recent patches introducing cpufreq_sched.c and gather feedback. Includes changes to scheduler and CPUfreq subsystem. If time permits, wander into the weeds regarding multi-scheduler class constraints (e.g. deadline scheduler bandwidth) and peripheral device constraints (e.g. <span class="caps">DDR</span> controller driver increasing <span class="caps">CPU</span> clock speed).</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Mike Turquette</b></p><div/><p>Linux kernel hacker since 2008, I focus on upstream power management frameworks. I currently maintain the clock framework. At Linaro I am involved in integrating CPUfreq with the scheduler.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/31532015-06-24T12:50:42-07:002015-06-29T08:20:30-07:00Framework for hardware accelerated virtio networking for NFVs2015-06-24T19:50:42Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Framework is aimed at addressing performance bottlenecks associated with Vhost-net networking using hardware offloads</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Virtio networking offers a flexible mechanism for sharing I/O interface among virtual machines. Despite the flexibility, there are performance challenges associated with virtio/vhost-net for virtual network functions. While handling high traffic rates, the vhost-net backend can become a bottleneck, leading to system saturation and in turn impacting virtual network function functionality.<br />
It’s possible to address the performance bottlenecks associated with vhost-net using hardware offloads.</p>
<p>In this talk we present a framework for leveraging hardware features like packet parsing, classification and distribution (zero copy Rx/Tx) for accelerating vhost-net networking. We would also like to share initial results from a proof of concept based on this framework. The framework tries to address the challenges associated with both north-south and east-west based traffic for network virtual functions.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Varun Sethi</b></p><div/><p>Varun Sethi is a Software Architect at Freescale Semiconductor and has been involved in virtualization software development for embedded Power Architecture SOCs. He has contributed to the <span class="caps">KVM</span> port for BookE.HV platforms and the e500mc core. He is the author for Freescale <span class="caps">PAMU</span> (<span class="caps">IOMMU</span>) driver. He’s working on enhancements for the <span class="caps">ARM</span> <span class="caps">SMMU</span> driver for Freescale Layerscape platforms. Varun is also looking at Virtio performance analysis and optimizations. He has made presentations on <span class="caps">KVM</span> and Virtio optimizations at <span class="caps">KVM</span> Forum (2012/2014). He has also conducted talks, on dependability in case of virtualized platforms at the <span class="caps">IEEE</span> <span class="caps">DSN</span> conference. He has a publication on <span class="caps">KVM</span> optimizations for BookE platforms without hardware virtualization assists, at <span class="caps">ASPLOS</span> 2013. He has conducted a joint talk on network virtualization at the Linux foundation collaboration summit.</p></li></ul></div></div><to_s/>