HTTP/1.1 -1 Read error in cache disk data: SuccessContent-Type: application/atom+xml
Last-Modified: Tue, 06 Mar 2018 21:58:13 GMT
ETag: "c97af-1cb15-566c589f9e740"
Content-length: 117525
Connection: Close
Proxy-Connection: Close
X-Cache: HIT from web2.osuosl.org
Server: ProxyTrack 0.5 (HTTrack 3.49.2)
tag:linuxplumbersconf.org,2005:/2017/ocw/proposalsLinux Plumbers Conference: Presentation Proposals2017-09-14T21:48:56-07:00tag:linuxplumbersconf.org,2005:Proposal/48372017-09-14T21:48:56-07:002017-09-15T09:16:11-07:00ktask: multithread cpu-intensive kernel work2017-09-15T04:48:56Z<div><dl><dt><b>One Line Summary:</b></dt><dd>As memory size and CPU counts increase on the largest systems, certain paths in the kernel that are single-threaded today are not scaling well and will become even bigger bottlenecks in the future. Examples include zeroing the biggest huge pages (e.g. 1G on x86), walking a process’s entire page table, and freeing large ranges of pages.
These paths have been well optimized in one thread, so now it's a matter of scaling up the CPUs to match the amount of work to be done.
One solution to this problem is ktask, a generic framework built on top of workqueues that takes care of splitting up a large task, starting a number of threads appropriate for the size of the task and the system, and load balancing the work between these threads. ktask uses internal limits to maintain an appropriate level of concurrency across the system.
I want to get feedback on the overall approach and discuss similar scalability issues others are having in the kernel.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>As memory size and <span class="caps">CPU</span> counts increase on the largest systems, certain paths in the kernel that are single-threaded today are not scaling well and will become even bigger bottlenecks in the future. Examples include zeroing the biggest huge pages (e.g. 1G on x86), walking a process’s entire page table, and freeing large ranges of pages.</p>
<p>These paths have been well optimized in one thread, so now it’s a matter of scaling up the CPUs to match the amount of work to be done.</p>
<p>One solution to this problem is ktask, a generic framework built on top of workqueues that takes care of splitting up a large task, starting a number of threads appropriate for the size of the task and the system, and load balancing the work between these threads. ktask uses internal limits to maintain an appropriate level of concurrency across the system.</p>
<p>I want to get feedback on the overall approach and discuss similar scalability issues others are having in the kernel.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>James E.J. Bottomley</b></p><div/><p>James Bottomley is a Distinguished Engineer at <span class="caps">IBM</span> Research where he<br />
works on Cloud and Container technology. He is also Linux Kernel<br />
maintainer of the <span class="caps">SCSI</span> subsystem. He has been a Director on the Board<br />
of the Linux Foundation and Chair of its Technical Advisory Board. He<br />
went to university at Cambridge for both his undergraduate and<br />
doctoral degrees after which he joined AT&T Bell labs to work on<br />
Distributed Lock Manager technology for clustering. In 2000 he helped<br />
found SteelEye Technology, a High availability company for Linux and<br />
Windows, becoming Vice President and <span class="caps">CTO</span>. He joined Novell in 2008 as<br />
a Distinguished Engineer at Novell’s <span class="caps">SUSE</span> Labs, Parallels (later Odin)<br />
in 2011 as <span class="caps">CTO</span> of Server Virtualization and <span class="caps">IBM</span> Research in 2016.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48362017-09-14T21:47:23-07:002017-09-15T09:14:13-07:00Automated testing of LKML patches with Clang2017-09-15T04:47:23Z<div><dl><dt><b>One Line Summary:</b></dt><dd>It's useful to have bots automatically test proposed patches and notify authors when they
fail to compile in certain configurations. Currently for Clang, a few folks interested in supporting it email
authors when patches break things, but regression finding is currently not automated for Clang. I'd like to
pick some brains as to what is a smart way to automatically test proposed patches for a seemingly limitless
number of kernel configurations.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>It’s useful to have bots automatically test proposed patches and notify authors when they<br />
fail to compile in certain configurations. Currently for Clang, a few folks interested in supporting it email<br />
authors when patches break things, but regression finding is currently not automated for Clang. I’d like to<br />
pick some brains as to what is a smart way to automatically test proposed patches for a seemingly limitless<br />
number of kernel configurations.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Nick Desaulniers</b></p><div/><p>Nick Desaulniers is a software engineer at Google working on the next generation of Nexus devices. Nick previously worked at Mozilla.</p>
<p>Contributing to Open Source software and an accessible Internet for all are some of the things that Nick is most passionate about.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48352017-09-14T21:37:56-07:002017-09-14T21:52:19-07:00Soft Affinity for Workloads2017-09-15T04:37:56Z<div><dl><dt><b>One Line Summary:</b></dt><dd>For weak performance isolation with multi-tenancy for some applications
by allowing space partitioning of CPUs, there is a need for applications
to specify soft affinity to some CPUs. So that one "group" does not
interfere with other "groups", yet allow the group to occupy the full
set of CPUs when no one else is using it. Currently, CFS allows time
sharing of CPUs and sched_setaffinity allows for "hard" affinity (or
restricting CPUs to a given mask or set).
One solution could be to to add a new cpumask other than cpus_allowed
(which is used for "hard" affinity). Let's say cpus_preferred. And allow
an application to set a preference to a set of CPUs if it so desires.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>For weak performance isolation with multi-tenancy for some applications<br />
by allowing space partitioning of CPUs, there is a need for applications<br />
to specify soft affinity to some CPUs. So that one “group” does not<br />
interfere with other “groups”, yet allow the group to occupy the full<br />
set of CPUs when no one else is using it. Currently, <span class="caps">CFS</span> allows time<br />
sharing of CPUs and sched_setaffinity allows for “hard” affinity (or<br />
restricting CPUs to a given mask or set).<br />
One solution could be to to add a new cpumask other than cpus_allowed<br />
(which is used for “hard” affinity). Let’s say cpus_preferred. And allow<br />
an application to set a preference to a set of CPUs if it so desires.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>James E.J. Bottomley</b></p><div/><p>James Bottomley is a Distinguished Engineer at <span class="caps">IBM</span> Research where he<br />
works on Cloud and Container technology. He is also Linux Kernel<br />
maintainer of the <span class="caps">SCSI</span> subsystem. He has been a Director on the Board<br />
of the Linux Foundation and Chair of its Technical Advisory Board. He<br />
went to university at Cambridge for both his undergraduate and<br />
doctoral degrees after which he joined AT&T Bell labs to work on<br />
Distributed Lock Manager technology for clustering. In 2000 he helped<br />
found SteelEye Technology, a High availability company for Linux and<br />
Windows, becoming Vice President and <span class="caps">CTO</span>. He joined Novell in 2008 as<br />
a Distinguished Engineer at Novell’s <span class="caps">SUSE</span> Labs, Parallels (later Odin)<br />
in 2011 as <span class="caps">CTO</span> of Server Virtualization and <span class="caps">IBM</span> Research in 2016.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48342017-09-14T21:35:41-07:002017-09-14T21:36:30-07:00Support for adding DT based thermal zones at runtime2017-09-15T04:35:41Z<div><dl><dt><b>One Line Summary:</b></dt><dd>During a recent project I encountered an issue where I needed support for thermal
zones that get instantiated by a devicetree overlay. In our specific case we had temperature sensors
on a daughtercard that is swap-able. I haven't found a good solution on how to make the current
of-thermal code work with runtime added thermal-zones, (assumption is all thermal zones present at boot)
and was wondering if people have clever ideas.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>During a recent project I encountered an issue where I needed support for thermal<br />
zones that get instantiated by a devicetree overlay. In our specific case we had temperature sensors<br />
on a daughtercard that is swap-able. I haven’t found a good solution on how to make the current<br />
of-thermal code work with runtime added thermal-zones, (assumption is all thermal zones present at boot)<br />
and was wondering if people have clever ideas.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>James E.J. Bottomley</b></p><div/><p>James Bottomley is a Distinguished Engineer at <span class="caps">IBM</span> Research where he<br />
works on Cloud and Container technology. He is also Linux Kernel<br />
maintainer of the <span class="caps">SCSI</span> subsystem. He has been a Director on the Board<br />
of the Linux Foundation and Chair of its Technical Advisory Board. He<br />
went to university at Cambridge for both his undergraduate and<br />
doctoral degrees after which he joined AT&T Bell labs to work on<br />
Distributed Lock Manager technology for clustering. In 2000 he helped<br />
found SteelEye Technology, a High availability company for Linux and<br />
Windows, becoming Vice President and <span class="caps">CTO</span>. He joined Novell in 2008 as<br />
a Distinguished Engineer at Novell’s <span class="caps">SUSE</span> Labs, Parallels (later Odin)<br />
in 2011 as <span class="caps">CTO</span> of Server Virtualization and <span class="caps">IBM</span> Research in 2016.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48332017-09-14T21:33:15-07:002017-09-15T09:13:35-07:00Test driven development (TDD) in the kernel2017-09-15T04:33:15Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Test driven development is a proven concept for user land
projects, and is as far as I have seen to some extent used for kernel development,
but is currently limited to what can be tested from user land, either via system call interfaces
or by creating mock environments to compile selected kernel code in user land.
Testing for instance kernel internal algorithms or usage of them, such as rbtrees, radix trees or sg lists cannot easily be
done this way. Also, once the test is in kernel code, there are some very interesting low hanging
opportunities such as for instance error injection and simple code coverage.
I'd like to demonstrate a tool I created for this purpose, and that we develop and use
internally in Oracle. It will be made available on Github for people to look at ahead of this
unconference. I submitted a talk proposal for this, but it did not reach up, but I modestly think
it is such a simple and good idea that I simply cannot wait yet another year without talking about it!</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Test driven development is a proven concept for user land <br />
projects, and is as far as I have seen to some extent used for kernel development,<br />
but is currently limited to what can be tested from user land, either via system call interfaces <br />
or by creating mock environments to compile selected kernel code in user land.<br />
Testing for instance kernel internal algorithms or usage of them, such as rbtrees, radix trees or sg lists cannot easily be <br />
done this way. Also, once the test is in kernel code, there are some very interesting low hanging <br />
opportunities such as for instance error injection and simple code coverage.<br />
I’d like to demonstrate a tool I created for this purpose, and that we develop and use<br />
internally in Oracle. It will be made available on Github for people to look at ahead of this <br />
unconference. I submitted a talk proposal for this, but it did not reach up, but I modestly think <br />
it is such a simple and good idea that I simply cannot wait yet another year without talking about it!</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Knut Omang</b></p><div/><p>Knut Omang has had a relationship with Unix and Linux since the mid-eighties. He wrote a Ph.D and collaborated on a driver for <span class="caps">SCI</span> (Scalable Coherent Interface) based I/O adapters, an early predecessor of today’s <span class="caps">RDMA</span> devices. For the last years he was the lead developer for the driver for an Infiniband <span class="caps">HCA</span> developed by Oracle.<br />
He currently works for the Oracle Linux Kernel networking group.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48322017-09-14T10:20:28-07:002017-09-14T10:25:35-07:00A unit test framework for the Linux kernel2017-09-14T17:20:28Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A unit test framework for the Linux kernel</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>This talk will present and demonstrate a simple and nonintrusive framework <br />
for writing and selectively running unit tests on external and internal <br />
kernel APIs.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Dhaval Giani</b></p><div/><p>Dhaval is currently one of the engineers at Oracle maintaining the Unbreakable Enterprise Kernel.</p>
<p>He has acquired a lot of interests over the years, catch him at the conference, and you might find some topic that interests both of you :-).</p></li><li><p><b>Knut Omang</b></p><div/><p>Knut Omang has had a relationship with Unix and Linux since the mid-eighties. He wrote a Ph.D and collaborated on a driver for <span class="caps">SCI</span> (Scalable Coherent Interface) based I/O adapters, an early predecessor of today’s <span class="caps">RDMA</span> devices. For the last years he was the lead developer for the driver for an Infiniband <span class="caps">HCA</span> developed by Oracle.<br />
He currently works for the Oracle Linux Kernel networking group.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48312017-09-13T13:18:10-07:002017-09-25T23:46:16-07:00Zero-copy Receive for virtio-net/vhost devices2017-09-13T20:18:10Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Leverage receive-side steering abilities of modern high speed network cards to implement zero-copy receive for virtio-net and vhost.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>In para-virtual networking with virtio-net/vhost, the copying of packets between the hypervisor and the guest is one of the major sources of overhead, especially for large packets. Although zero-copy transmit was merged into the Linux kernel a few years ago, the “receive side zero copy” item is still in the <span class="caps">KVM</span> NetworkingTodo. This BoF discusses issues in implementing zero-copy receive for virtio-net and vhost that leverages receive-side steering abilities of modern high speed network cards.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Kalman Meth</b></p><div/><p>Dr. Kalman Meth is an architect in the Cloud Platforms department at the <span class="caps">IBM</span> Haifa Research Labs (<span class="caps">HRL</span>). Kalman is one of the authors of the iSCSI protocol specification (RFCs 3720, 7143) and is active in various systems and storage related activities. He received his Masters and PhD degrees from Courant Institute at New York University, while specializing in numerical analysis and computational fluid dynamics. His interests include: operating systems, file systems, storage systems, distributed and parallel computing, real-time computing, and numerical analysis.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48302017-09-13T12:06:51-07:002017-09-14T10:25:38-07:00Unit Testing and Mocking2017-09-13T19:06:51Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion on the possible value of adding unit tests to the kernel.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Currently most testing in the kernel is done on a booted kernel and is primarily driven by the userland. Some portions of the kernel may be well tested, but in practice many code paths in the kernel are not exercised before code is submitted. Unit testing and mocking is one possible way to make it much easier to exercise arbitrary code paths.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Brendan Higgins</b></p><div/><p>Brendan is currently a software engineer on the server software and hardware team, known as Platforms, and works primarily on <span class="caps">BMC</span> development and actively contributes to the OpenBMC project and the Linux kernel.</p>
<p>His current focus is on interfacing to the <span class="caps">BMC</span> via <span class="caps">IPMI</span>, and <span class="caps">MCTP</span>. He is also interested in low speed peripherals such as <span class="caps">SPI</span>, <span class="caps">LPC</span>, eSPI and I2C.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48292017-09-13T05:50:50-07:002017-09-13T05:51:51-07:00Tying TPMs into the EFI boot stub2017-09-13T12:50:50Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Tying TPMs into the EFI boot stub</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Tying TPMs into the <span class="caps">EFI</span> boot stub</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 Google, focusing on Linux desktop and infrastructural security.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48272017-09-13T05:44:03-07:002017-09-13T05:44:23-07:00Introduction2017-09-13T12:44:03Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Introduction</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Introduction</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 Google, focusing on Linux desktop and infrastructural security.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48262017-09-12T13:30:03-07:002017-09-12T15:06:35-07:00syzkaller status update2017-09-12T20:30:03Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A quick update on syzkaller</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Syzkaller is a coverage-guided system call fuzzer for the Linux kernel. We give a quick update on what has been done in the last year and what are the directions for the future work.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Alexander Potapenko</b></p><div/><p>I’ve been involved with dynamic tools for memory detection since 2009.<br />
Have experience with hacking Valgrind, <span class="caps">LLVM</span>, Google Perftools on both Linux and Mac OS X.<br />
Co-author of AddressSanitizer, the state of the art addressability error detector.<br />
Right now I’m contributing to KernelAddressSanitizer and developing KernelMemorySanitizer.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48252017-09-12T11:47:31-07:002017-09-15T09:06:45-07:00KernelMemorySanitizer against uninitialized memory2017-09-12T18:47:31Z<div><dl><dt><b>One Line Summary:</b></dt><dd>I'll present KernelMemorySanitizer, a new tool for detecting uses of uninitialized values in the kernel.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>KernelMemorySanitizer is a new tool that detects uses of uninitialized values in the Linux kernel.<br />
The tool is based on compiler instrumentation and is times faster than kmemcheck, so one can use it with fuzzers or other load.<br />
I’ll outline some implementation details and the possible applications of the same approach to other problems in the kernel.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Alexander Potapenko</b></p><div/><p>I’ve been involved with dynamic tools for memory detection since 2009.<br />
Have experience with hacking Valgrind, <span class="caps">LLVM</span>, Google Perftools on both Linux and Mac OS X.<br />
Co-author of AddressSanitizer, the state of the art addressability error detector.<br />
Right now I’m contributing to KernelAddressSanitizer and developing KernelMemorySanitizer.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48242017-09-11T10:00:54-07:002017-09-12T15:07:33-07:00Kernel ABI2017-09-11T17:00:54Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Kernel ABI</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Last year we discussed creating a specification for the kernel’s <span class="caps">ABI</span>.</p>
<p>We have a working skeleton, but are missing the meat – actual syscall and ioctl definitions (and the skeleton could use some work too).</p>
<p>Let’s discuss that for a bit…</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Sasha Levin</b></p><div/><p>Sasha is the maintainer of the Native Linux <span class="caps">KVM</span> Tools, a from-scratch implementation of a <span class="caps">KVM</span> hypervisor designed to live inside the kernel tree. Sasha is currently employed by Oracle, working in the Ksplice group which provides rebootless security updates to the Linux kernel. Previously he worked in Host Dynamics which provided the ability to dynamically shape the size of virtual guests according to their needs in a given moment. Before that, he worked in the Israeli army’s cryptography unit.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48232017-09-11T09:58:46-07:002017-09-12T15:08:09-07:00Update on Distro/Stable testing2017-09-11T16:58:46Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Update from the BoF</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>A summary of the discussions during the BoF and discussions arising from that.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Dhaval Giani</b></p><div/><p>Dhaval is currently one of the engineers at Oracle maintaining the Unbreakable Enterprise Kernel.</p>
<p>He has acquired a lot of interests over the years, catch him at the conference, and you might find some topic that interests both of you :-).</p></li><li><p><b>Sasha Levin</b></p><div/><p>Sasha is the maintainer of the Native Linux <span class="caps">KVM</span> Tools, a from-scratch implementation of a <span class="caps">KVM</span> hypervisor designed to live inside the kernel tree. Sasha is currently employed by Oracle, working in the Ksplice group which provides rebootless security updates to the Linux kernel. Previously he worked in Host Dynamics which provided the ability to dynamically shape the size of virtual guests according to their needs in a given moment. Before that, he worked in the Israeli army’s cryptography unit.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48222017-09-11T09:57:37-07:002017-09-12T15:06:15-07:00Fuzzers Panel2017-09-11T16:57:37Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Fuzzers Panel</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Get the maintainers of the various Fuzzers to talk about where fuzzing is going</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Dhaval Giani</b></p><div/><p>Dhaval is currently one of the engineers at Oracle maintaining the Unbreakable Enterprise Kernel.</p>
<p>He has acquired a lot of interests over the years, catch him at the conference, and you might find some topic that interests both of you :-).</p></li><li><p><b>Sasha Levin</b></p><div/><p>Sasha is the maintainer of the Native Linux <span class="caps">KVM</span> Tools, a from-scratch implementation of a <span class="caps">KVM</span> hypervisor designed to live inside the kernel tree. Sasha is currently employed by Oracle, working in the Ksplice group which provides rebootless security updates to the Linux kernel. Previously he worked in Host Dynamics which provided the ability to dynamically shape the size of virtual guests according to their needs in a given moment. Before that, he worked in the Israeli army’s cryptography unit.</p></li><li><p><b>Dave Jones</b></p><div/><p>Former kernel developer at Red Hat, now production engineer at Facebook.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48212017-09-10T20:57:46-07:002017-09-12T15:07:55-07:00ktest - How is it used and what more can be done2017-09-11T03:57:46Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion on ways to improve ktest and how it is used</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>ktest.pl has been in the kernel since 2.6.38. It is a testing framework to automate various testing. It can test patch sets, individual builds, bisecting and more. The focus of this discussion will be on how people currently use ktest.pl and what more can be added to it.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Steven Rostedt</b></p><div/><p>Steven Rostedt has been working with the Linux kernel since 2001. He currently works for Red Hat working in their Messaging Real-time Grid (<span class="caps">MRG</span>) division. He created and maintains Ftrace, the official Linux kernel tracer, and is the current real-time kernel stable maintainer.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48202017-09-09T20:30:29-07:002017-09-09T21:59:07-07:00Debugging BPF programs and infrastructure remotely on live systems2017-09-10T03:30:29Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Share best practices for debugging BPF programs in production systems and discuss possible improvements.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p><span class="caps">BPF</span> has gained many new features over the past couple years and is now being deployed on production systems. But, being a relatively new subsystem debugging <span class="caps">BPF</span> programs and infrastructure has gotten less of a focus. For projects, such as Cilium, which use <span class="caps">BPF</span> as a core component in the networking stack remote debugging is starting to become a requirement.</p>
<p>In this session we would like to share some ideas we have for how to remotely debug <span class="caps">BPF</span> infrastructure and programs on live systems. As well as hopefully come up with new and even better techniques.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>John Fastabend</b></p><div/><p>John is a software engineer.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48192017-09-09T02:05:21-07:002017-09-09T07:14:05-07:00Safe multicore scheduling in a Linux cluster environment2017-09-09T09:05:21Z<div><dl><dt><b>One Line Summary:</b></dt><dd>The Linux scheduler has become fairly complex and is not always able to preserve work-conservation. We present Ipanema, a DSL to write simple and safe schedulers with proven properties.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Modern clusters rely on servers containing dozens of cores that have high infrastructure cost and energy consumption. Therefore, it is economically essential to exploit their full potential at both the application and the system level. While today’s clusters typically rely on Linux, recent research shows that the Linux scheduler suffers from performance bugs that lead to core under-usage. The consequences are wasted energy, bad infrastructure usage, and lower service response time.</p>
<p>The fundamental source of such performance bugs is that the Linux scheduler, being monolithic, has become too complex. In this talk, we present ongoing work on the Ipanema project that proposes to switch to a set of simple schedulers, each tailored to a specific application. Our vision raises scientific challenges in terms of easily developing schedulers and proving them safe. The key to our approach is designing a Domain-Specific Language for multicore kernel scheduling policy development, associated verification tools and a Linux runtime system.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Jean-Pierre Lozi</b></p><div/><p>Jean-Pierre Lozi is a Principal Member of the Technical Staff at Oracle Labs Zurich, working on distributed graph processing (<span class="caps">PGX</span>.D). He was formerly an Associate Professor at the University of Nice Sophia Antipolis in France, where he was working on multicore architectures, lock and synchronization algorithms, and OS schedulers.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48182017-09-08T14:51:51-07:002017-09-14T11:37:17-07:00TPM2 Software Stack: Status Report and Future Work2017-09-08T21:51:51Z<div><dl><dt><b>One Line Summary:</b></dt><dd>TPM2 Software Stack: Retrospective and Future Work</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>This talk will describe the last year of development in Intel’s implementation of the <span class="caps">TCG</span> TPM2 Software Stack specifications and the work we have planned for the upcoming year. We’ll cover everything from community building, our new test harness, our continuous integration system to language bindings and integrating the APIs into event driven application environments like GLib.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>philip tricca</b></p><div/><p>Philip is a platform architect in Intel’s platform security division working to enable use of the Trusted Platform Module (<span class="caps">TPM</span>) and <span class="caps">SGX</span> in open source. Recently Phil has taken over maintainership of Intel’s implementation of the TPM2 software stack and has been obsessing over system integrity and measurement architectures for years. In his spare time he maintains the meta-measured Open Embedded meta layer where he brings together the various integrity measurement components from TPM2 to Grub2 <span class="caps">SRTM</span> all the way up to the userspace infrastructure. Additionally he’s a periodic contributor to meta-selinux and the OpenXT project.</p></li><li><p><b>Jarkko Sakkinen</b></p><div/><p>Software Engineer @ Intel</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48172017-09-08T10:09:44-07:002017-09-13T05:51:41-07:00Crypto System Integration for TPM 2.02017-09-08T17:09:44Z<div><dl><dt><b>One Line Summary:</b></dt><dd>status and next steps for crypto system integration of TPM 2.0</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>One of the most useful functions of the <span class="caps">TPM</span> would be correct integration with the crypto systems. The first issue to solve before doing this is the form of the <span class="caps">TPM</span> key, which should really be standardised.</p>
<p>Based on this, I’ll show integrations for openssl (with gnutls easily done as well), gnome-keyring via pkcs11 and possibly openssl.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>James E.J. Bottomley</b></p><div/><p>James Bottomley is a Distinguished Engineer at <span class="caps">IBM</span> Research where he<br />
works on Cloud and Container technology. He is also Linux Kernel<br />
maintainer of the <span class="caps">SCSI</span> subsystem. He has been a Director on the Board<br />
of the Linux Foundation and Chair of its Technical Advisory Board. He<br />
went to university at Cambridge for both his undergraduate and<br />
doctoral degrees after which he joined AT&T Bell labs to work on<br />
Distributed Lock Manager technology for clustering. In 2000 he helped<br />
found SteelEye Technology, a High availability company for Linux and<br />
Windows, becoming Vice President and <span class="caps">CTO</span>. He joined Novell in 2008 as<br />
a Distinguished Engineer at Novell’s <span class="caps">SUSE</span> Labs, Parallels (later Odin)<br />
in 2011 as <span class="caps">CTO</span> of Server Virtualization and <span class="caps">IBM</span> Research in 2016.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48162017-09-07T17:37:07-07:002017-09-14T14:24:49-07:00BoF - Open Discussion2017-09-08T00:37:07Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Open Discussion for issues not covered</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Open Discussion for issues not covered or continuation of discussions not finished</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>John Stultz</b></p><div/><p>John Stultz, working with Linaro, has for a number of years been focused on getting Android functionality upstream into the mainline kernel.</p>
<p>He has also recently collaborated with others to get the 96boards HiKey board in <span class="caps">AOSP</span> and working to get its functionality upstream as well.</p>
<p>Previously, he worked on x86 server enablement, and Enterprise Realtime Linux.</p>
<p>He’s also a maintainer of the Linux timekeeping subsystem.</p></li><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><li><p><b>Todd Kjos</b></p><div/><p>Todd Kjos is on the Android systems team at Google focusing on performance and power. He has recently been optimizing the Energy Aware Scheduler for Android workloads.</p>
<p>Previously he worked on x86 and Itanium server enablement and Itanium virtualization.</p></li><li><p><b>Serban Constantinescu</b></p><div/><p>I am Software Engineer working for <span class="caps">ARM</span> as part of the Android team. In the past few years I have been involved on a couple of Android subsystems from Ashmem and Binder to libc, Dalvik and <span class="caps">ART</span>. I have been actively involved in the ARM64 Android bootstrap and previously I have been maintaining the <span class="caps">ARM</span> Versatile Express Android ports.</p>
<p>At roots I have an electronics background, graduated the Polytechnic University of Bucharest and a master at Bristol University.</p>
<p>Ask me about:</p>
<ul>
<li><span class="caps">ARM</span></li>
<li>Binder</li>
<li>Bionic</li>
<li>Art</li>
<li>Android</li>
<li>OpenCV (though a bit rusty now)</li>
</ul>
<p>You can check my latest patches here:</p>
<ul>
<li><span class="caps">AOSP</span>: <a href="http://goo.gl/s5XLS2" rel="nofollow">http://goo.gl/s5XLS2</a></li>
<li>Linux Kernel: <a href="http://goo.gl/cHigeb" rel="nofollow">http://goo.gl/cHigeb</a></li>
</ul></li><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/48152017-09-07T12:13:46-07:002017-09-11T07:38:14-07:00BIOS Manageability discussion2017-09-07T19:13:46Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion around how to present interfaces that manage BIOS settings to the kernel and userspace</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>After a heated discussion on <span class="caps">LKML</span> a few months ago, this is to discuss more about what an interface that can manage <span class="caps">BIOS</span> setup options can look like.</p>
<p>Some manufacturers are pushing for cross-OS industry standard manageability interfaces with the goals of having parity at the options that can be manipulated without compromising security.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Mario Limonciello</b></p><div/><p>Mario is a software architect at Dell.</p>
<p>He is involved in all Linux OSes that are shipped with Dell client systems.</p>
<p>He also spends a lot of time working on technology related to flashing firmware and <span class="caps">BIOS</span> interfaces on Dell systems.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48142017-09-07T10:16:54-07:002017-09-08T17:27:11-07:00Using runtime PM callbacks for system suspend/resume in PM core2017-09-07T17:16:54Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion on a recent proposal to change the PM core so that it can reuse runtime PM callbacks directly for system suspend/resume in some situations.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The integration of the system suspend/resume infrastructure with the runtime PM framework has been a popular topic over the last few years, but there still is quite a lot to do in that area. The problems that have come up most recently are the lack of a sufficiently clean way to reuse runtime PM callbacks for system suspend/resume in some configurations (e.g. if the <span class="caps">ACPI</span> PM domain is used) and suboptimal behavior in some cases when devices are already runtime-suspended during system resume and may stay suspended or when devices suspended for system sleep may stay suspended (in runtime suspend) after system wakeup.</p>
<p>The latest idea on how to deal with these issues is to allow drivers to set flags informing the PM core on the driver capabilities and/or preferences and modify the PM core (and possibly middle-layer code) to adjust its behavior in accordance with the driver flags.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Rafael Wysocki</b></p><div/><p>Rafael maintains the Linux kernel’s core <span class="caps">ACPI</span> and power management code, including the core infrastructure for IO device PM, <span class="caps">CPU</span> PM and system suspend/hibernation. He works at Intel Open Source Technology Center as a Software Engineer with focus on the mainline Linux kernel. He has been actively contributing to Linux since 2005, in particular to the kernel’s suspend/hibernate subsystem, power management in general (IO runtime PM framework, cpufreq, cpuidle, PM QoS, wakeup framework etc.), hot-plug infrastructure, <span class="caps">ACPI</span> core and <span class="caps">PCI</span> core. Since 2008 he has participated in multiple Linux Foundation conferences and other Linux-related events, including the Linux Foundation Collaboration Summit, LinuxCon (North America/Japan/Europe), Linux Plumbers Conference, Linux.conf.au, LinuxTag, and Ottawa Linux Symposium. He holds a PhD in physics from the University of Warsaw, Poland (2002).</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48132017-09-07T09:33:03-07:002017-09-14T20:41:59-07:00membarrier: core serialization for reclaim of JIT code memory2017-09-07T16:33:03Z<div><dl><dt><b>One Line Summary:</b></dt><dd>We need input from architecture maintainers about the best way to proceed to guarantee core serialization instructions within the membarrier system call (and possibly within the scheduler).</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>(Issue resolved in Hallway track due to unscheduled resolution of problem)</p>
<p>This is an attempt to gather scheduler maintainers, and architecture maintainers, in the same room to get feedback on core serialization guarantees from each architecture when the scheduler is executed, and the requirements each architecture has with respect to core serialization in situations of reclaim of <span class="caps">JIT</span> code.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Mathieu Desnoyers</b></p><div/><p>Mathieu Desnoyers main contributions are in the area of tracing (monitoring/performance analysis/debugging) and scalability, both at the kernel and user-space levels. He maintains the LTTng project, the Userspace <span class="caps">RCU</span> library, as well as the Linux kernel membarrier system call. He works in close collaboration with the telecommunication industry, many Linux distributions, and with customers developing hardware scaling from small embedded devices to<br />
large-deployment servers. He is <span class="caps">CEO</span> and Senior Software Architect at EfficiOS.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48122017-09-07T04:38:18-07:002017-09-18T03:01:53-07:00SVM on ARM SMMUv32017-09-07T11:38:18Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A brief overview of SVM on ARM.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Earlier this year I proposed an implementation of <span class="caps">ATS</span>, <span class="caps">PRI</span> and <span class="caps">PASID</span> in the <span class="caps">ARM</span> SMMUv3 driver. Along with that, I proposed to consolidate the <span class="caps">API</span> for host <span class="caps">SVM</span>, and add an interface to <span class="caps">VFIO</span> for userspace drivers.</p>
<p>In this session I’d like to talk about the differences between <span class="caps">ARM</span> <span class="caps">SMMU</span> and other architectures with regard to <span class="caps">SVM</span>, and discuss the requirements of the host <span class="caps">SVM</span> interface.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Jean-Philippe Brucker</b></p><div/><p>Software engineer at <span class="caps">ARM</span>, currently focusing on <span class="caps">IOMMU</span> virtualization.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48112017-09-07T03:54:01-07:002017-09-18T02:59:26-07:00virtio-iommu: a paravirtualized IOMMU2017-09-07T10:54:01Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Status report on virtio-iommu, introduction of extensions for vSVM: page table handover, PASID and fault reporting.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Virtio-iommu is a new <span class="caps">IOMMU</span> architecture introduced in early 2017, that focuses on virtualization.</p>
<p>Being paravirtualized and relying on an existing transport mechanism, it is efficient and easy to implement. It is designed to integrate seamlessly with software APIs such as <span class="caps">VFIO</span>, removing the need for page table emulation.</p>
<p>I will briefly describe how it works at the moment. Then I’ll discuss future extensions for <span class="caps">SVM</span> virtualization, how they fit with current vSVM work and the few challenges that stem from using hardware acceleration in a multi-arch <span class="caps">IOMMU</span>.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Jean-Philippe Brucker</b></p><div/><p>Software engineer at <span class="caps">ARM</span>, currently focusing on <span class="caps">IOMMU</span> virtualization.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48102017-09-06T23:28:47-07:002017-09-07T17:32:40-07:00The Android kernel lowmemorykiller replacement2017-09-07T06:28:47Z<div><dl><dt><b>One Line Summary:</b></dt><dd>The lowmemorykiller was dropped from staging, so this will cover what will replace it.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The Android lowmemorykiller has been around for ahwile, it was added to staging, and work was done with memcg groups to allow a userland lowmemorykiller daemon to be used instead, but the userspace implementation never performed as well as the in-kernel implementation. Recently the lowmemorykiller driver was dropped from staging, so this talk will discuss what will be used to replace it.</p></dl><div><p><b>Speakers:</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><li><p><b>Todd Kjos</b></p><div/><p>Todd Kjos is on the Android systems team at Google focusing on performance and power. He has recently been optimizing the Energy Aware Scheduler for Android workloads.</p>
<p>Previously he worked on x86 and Itanium server enablement and Itanium virtualization.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48092017-09-06T20:16:20-07:002017-09-11T21:35:14-07:00NVMe Surprise Removal2017-09-07T03:16:20Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion on problems related to supporting NVMe surprise removal</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Over the years surprise drive removal has been solved multiple times for various storage stacks like <span class="caps">SCSI</span>, <span class="caps">SATA</span>, <span class="caps">SAS</span> as a collective community effort involving drive vendors, OEMs, OS vendors and kernel developers. This problem has largely remained unsolved for NVMe flash controllers.</p>
<p>Recent work like <a href="https://www.spinics.net/lists/linux-pci/msg58123.html" rel="nofollow">https://www.spinics.net/lists/linux-pci/msg58123.html</a> have been welcome changes towards solving this problem in the kernel <span class="caps">PCI</span> subsystem.<br />
However, there are many other scenarios that can break the system on a NVMe drive surprise removal.</p>
<p>In this talk and discussion we bring to fore the challenges, test case scenarios and related stack problems that impede support for NVMe drive surprise removal. We also discuss possible solutions and steps the community as a whole could take to solve this problem. Solving this problem will go a long way in plumbing the storage stack for NVMe devices.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Shyamkumar Iyer</b></p><div/><p>Shyam Iyer is a Sr. Principal Software Sys Engineer in Dell EMC’s Server Solutions Office of the <span class="caps">CTO</span> focused on enabling technology concepts and ecosystem evangelization. Shyam is a software technologist with a background working with solutions in Data Center Environments and has been involved with Linux kernel development in the past as part of the Dell Linux Engineering team.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48082017-09-06T17:38:26-07:002017-09-07T18:56:16-07:00new kprobe api2017-09-07T00:38:25Z<div><dl><dt><b>One Line Summary:</b></dt><dd>lessons from using kprobe text based api and proposed extensions</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>As a heavy user of kprobe, uprobe, tracepoints together with bpf tracing, we’ve hit a number of issues with text (tracefs) based kprobe api that led us to propose new binary (file descriptor based) api to address them</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Alexei Starovoitov</b></p><div/><p>software engineer at facebook</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48072017-09-06T15:17:10-07:002017-09-14T08:48:52-07:00Treble and Kernel2017-09-06T22:17:10Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Discussion around changes made to Android and to the tests / requirements as part of Project Treble</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Discussion around changes made to Android and to the tests / requirements as part of Project Treble</p>
- First stage mount
- New <span class="caps">VTS</span> Tests for kernel compatibility
- Recommendations for maintaining SoC specific kernels to address the kernel fragmentation seen across the Android ecosystem.
- Minimum Kernel Version</dl><div><p><b>Speaker:</b></p><ul><li><p><b>Sandeep Patil</b></p><div/><p>Staff software engineer at Google. Worked on Project Ara and now working on Project Treble.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48052017-09-06T13:45:26-07:002017-09-14T06:31:37-07:00Testing Android common and LTS kernels2017-09-06T20:45:26Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A quick overview of Android common and LTS testing efforts.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>In this session Marissa will talk about the test coverage of vanilla Android common kernels, their plan to catch regressions with each new patch before it can be merged.</p>
<p>Amit will provide a quick overview of linux kernel functional testing efforts at Linaro, test suites (<span class="caps">LTP</span>, kselftests, hugetlbfs, Android <span class="caps">VTS</span> etc) currently deployed in the lab, pain points, future plans and how can you join/help.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Amit Pundir</b></p><div/><p>Android Engineer at Linaro. <span class="caps">AOSP</span> contributor.</p>
<p>Maintain Android kernel topic branches for linaro-stable and mainline tracking tree at Linaro. Contributing to Stable kernels lately.</p>
<p>Previously worked on Indian language computing projects, and also did fair bit of <span class="caps">GUI</span> development (Clutter/Qt) on OMAP3/Beagle-family boards.</p></li><li><p><b>Marissa Wall</b></p><div/><p>Marissa Wall is a software engineer on the Android systems team at Google. She recently joined the team and began working on Android display.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48042017-09-06T11:44:13-07:002017-09-06T11:44:55-07:00Tracing and Kernel ABI2017-09-06T18:44:13Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Talk about how we want to deal with trace points as a user space interface</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Have a discussion about how important it is to maintain consistent trace points and how to deal with changing them as the code in the kernel changes around them.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Josef Bacik</b></p><div/><p>Linux kernel developer working for Facebook. I’m a maintainer for Btrfs, and work on file system and tracing related things.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48032017-09-06T06:55:46-07:002017-09-12T14:35:09-07:00PCIe error statistics reporting2017-09-06T13:55:46Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Report PCIe error statistics via sysfs</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p><span class="caps">PCI</span> Express Root Port Advanced Error Reporting currently only supports reporting errors to the console. Having to monitor console logs is tedious and complicated when running a large data center.</p>
<p>We would like to propose adding software counters to the <span class="caps">PCIAER</span> driver and make these accessible via sysfs.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Jes Sorensen</b></p><div/><p>Jes Sorensen is a Trained Monkey on the kernel team at Facebook, where he primarily works on Software <span class="caps">RAID</span> and networking drivers. In addition he amuses himself by organizing Linux conferences and writing Wi-Fi device drivers when he has the time for it.</p>
<p>Jes Sorensen has worked on the Linux kernel and userland since 1993. He has worked on a number of device drivers, <span class="caps">RAID</span>, <span class="caps">KVM</span>, high speed networking, Linux/ia64, Linux/m68k, the system libraries (glibc), and high-end <span class="caps">NUMA</span> systems.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48022017-09-06T03:19:19-07:002017-09-21T08:21:06-07:00EFI + Intel TXT and TPM + Xen/Linux - how to make it work2017-09-06T10:19:19Z<div><dl><dt><b>One Line Summary:</b></dt><dd>EFI + Intel TXT and TPM + Xen/Linux - how to make it work</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>After some investigation it looks that <span class="caps">EFI</span><br />
+ tboot + Xen does not work. The problem is that<br />
tboot treats <span class="caps">EFI</span> as untrusted stuff and shuts down<br />
all services. However, these services are needed<br />
to boot Xen properly. So, this tboot behavior<br />
makes it completely unusable with Xen. Linux is<br />
hit by this issue, too. It is less severe because<br />
it boots but due to a lack of <span class="caps">EFI</span> runtime services<br />
it is not possible to run e.g. efibootmgr which<br />
manages machine boot config. Hence, this means<br />
that we should hammer out proper approach to that<br />
problem. At the beginning of discussion we should<br />
review <span class="caps">EFI</span> infrastructure security. This should<br />
lead to a decision about <span class="caps">EFI</span> availability in<br />
measured environments. If yes, then we should<br />
decide what and how should be exposed. It is also<br />
worth considering here solutions providing<br />
functionality similar to tboot, e.g. TrustedGRUB,<br />
<span class="caps">EFI</span> <span class="caps">TBOOT</span>, etc.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Daniel Kiper</b></p><div/><p>Daniel Kiper works as software developer<br />
for Oracle. He is responsible for Xen boot<br />
code development. He also played with GRUB2<br />
and due to that last year he was appointed<br />
as one of the <span class="caps">GRUB</span> maintainers. Earlier he<br />
worked on kexec, kdump, makedumpfile, crash<br />
tool and memory hotplug development.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/48012017-09-05T20:52:39-07:002017-09-06T14:04:06-07:00Intro2017-09-06T03:52:39Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Introduction to the Android Microconference</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Quick introduction and overview of the microconference.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>John Stultz</b></p><div/><p>John Stultz, working with Linaro, has for a number of years been focused on getting Android functionality upstream into the mainline kernel.</p>
<p>He has also recently collaborated with others to get the 96boards HiKey board in <span class="caps">AOSP</span> and working to get its functionality upstream as well.</p>
<p>Previously, he worked on x86 server enablement, and Enterprise Realtime Linux.</p>
<p>He’s also a maintainer of the Linux timekeeping subsystem.</p></li><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><li><p><b>Todd Kjos</b></p><div/><p>Todd Kjos is on the Android systems team at Google focusing on performance and power. He has recently been optimizing the Energy Aware Scheduler for Android workloads.</p>
<p>Previously he worked on x86 and Itanium server enablement and Itanium virtualization.</p></li><li><p><b>Serban Constantinescu</b></p><div/><p>I am Software Engineer working for <span class="caps">ARM</span> as part of the Android team. In the past few years I have been involved on a couple of Android subsystems from Ashmem and Binder to libc, Dalvik and <span class="caps">ART</span>. I have been actively involved in the ARM64 Android bootstrap and previously I have been maintaining the <span class="caps">ARM</span> Versatile Express Android ports.</p>
<p>At roots I have an electronics background, graduated the Polytechnic University of Bucharest and a master at Bristol University.</p>
<p>Ask me about:</p>
<ul>
<li><span class="caps">ARM</span></li>
<li>Binder</li>
<li>Bionic</li>
<li>Art</li>
<li>Android</li>
<li>OpenCV (though a bit rusty now)</li>
</ul>
<p>You can check my latest patches here:</p>
<ul>
<li><span class="caps">AOSP</span>: <a href="http://goo.gl/s5XLS2" rel="nofollow">http://goo.gl/s5XLS2</a></li>
<li>Linux Kernel: <a href="http://goo.gl/cHigeb" rel="nofollow">http://goo.gl/cHigeb</a></li>
</ul></li><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/48002017-09-05T18:28:40-07:002017-09-14T14:24:39-07:00SDCardFS Upstreaming2017-09-06T01:28:40Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A description of SDCardfs and the plans to upstream it.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Android has recently switched from using <span class="caps">FUSE</span> for its emulated sdcard layer to using SDCardfs. This talk will talk about that transition, and the future plans to upstream the replacement.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Daniel Rosenberg</b></p><div/><p>Daniel Rosenberg is a software engineer on the Android systems team at Google focusing on filesystems.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47992017-09-05T16:51:05-07:002017-09-13T09:42:49-07:00Clang built kernels2017-09-05T23:51:05Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A talk on compiling the Linux kernel with Clang/LLVM.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>In 2017 we started recommending to Android vendors and OEMs that we’d like to transition Android’s toolchain to be entirely Clang/<span class="caps">LLVM</span> based. We plan to hold a discussion on what this means and looks like for building the Linux kernel. Our patch-set, common issues, and benefits will be reviewed.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Nick Desaulniers</b></p><div/><p>Nick Desaulniers is a software engineer at Google working on the next generation of Nexus devices. Nick previously worked at Mozilla.</p>
<p>Contributing to Open Source software and an accessible Internet for all are some of the things that Nick is most passionate about.</p></li><li><p><b>Greg Hackmann</b></p><div/><p>Greg Hackmann is a software engineer on the Android systems team at Google. Since joining Google in 2012, his work has focused on Linux kernel support for display hardware.</p>
<p>Greg earned a PhD in computer science from Washington University in St. Louis.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47982017-09-05T15:56:34-07:002017-09-13T05:46:58-07:00Trusted Boot on Linux2017-09-05T22:56:34Z<div><dl><dt><b>One Line Summary:</b></dt><dd>What's been done over the past year, and what's still needed?</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Trusted Boot on Linux has not been in a great state for years, but we’re finally getting the pieces together. This presentation will give an overview of where we are, along with the work that still needs to be done and how distributions can make use of it.</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 Google, focusing on Linux desktop and infrastructural security.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47972017-09-05T10:49:11-07:002017-09-13T10:07:40-07:00The Android Emulator and Upstream QEMU2017-09-05T17:49:11Z<div><dl><dt><b>One Line Summary:</b></dt><dd>A historical perspective and future outlook of the Android emulator and upstream QEMU and their differences.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>We will briefly cover the history of the Android emulator and its relationship to <span class="caps">QEMU</span>, as well as discuss what the Goldfish platform is. We will then discuss the challenge we had with the Android emulator when AAarch64 was introduced (64-bit <span class="caps">ARM</span>), and how Google had to make a non-trivial choice between backporting functionality from upstream <span class="caps">QEMU</span> or move to a newer <span class="caps">QEMU</span>. Google eventually changed the basis of their Android emulator to a (much) newer <span class="caps">QEMU</span> which included AArch64 support. We will briefly cover the differences between the Google Android emulator and upstream <span class="caps">QEMU</span> and what it would it would take to bridge the gap between these two code bases. We open up to discussions about interests in upstreaming more Android emulator support to the <span class="caps">QEMU</span> project or other interests and challenges the community has with the Android emulator today.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Christoffer Dall</b></p><div/><p>Christoffer Dall is the Virutalization Tech Lead in Linaro, overseeing a team of <span class="caps">QEMU</span> and <span class="caps">KVM</span> engineers with a focus on <span class="caps">ARM</span>. Christoffer also co-maintains <span class="caps">KVM</span>/<span class="caps">ARM</span> in Linux, and has worked on the AArch64 Android Emulator support in Linux and <span class="caps">QEMU</span>. Christoffer has a PhD in computer science from Columbia University, specializing in operating systems and virtualization, and has industry experience from VMware and and various other companies.</p></li><li><p><b>Jin Qian</b></p><div/><p>Jin Qian is a software engineer on the Android systems team at Google.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47962017-09-05T10:31:41-07:002017-09-12T22:20:48-07:00LTS Efforts2017-09-05T17:31:41Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Focused Efforts around LTS</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>We will talk about focused efforts around <span class="caps">LTS</span> in the context of Android Kernels – why do we need them, what are we doing, and how.</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’s 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>Sumit previously 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/47952017-09-04T13:53:39-07:002017-09-20T12:39:34-07:00Privileged actions in unprivileged containers2017-09-04T20:53:39Z<div><dl><dt><b>One Line Summary:</b></dt><dd>How to selectively allow privileged actions from otherwise unprivileged containers?</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Unprivileged containers, that is, containers which use a user namespace to map their UIDs and GIDs to an unprivileged range, have very limited kernel privileges.</p>
<p>Those privileges are no higher than what a normal unprivileged user would have on the system and are restricted to the container’s namespaces.<br />
Anything which isn’t owned by the container and isn’t part of a namespace will typically be rejected by the kernel.</p>
<p>Some common examples include:<br />
– Creating device nodes<br />
– Mounting filesystems<br />
– Setting up loop devices<br />
– Raising a process nice level<br />
– Raising the <span class="caps">OOM</span> score of a process</p>
<p>In this presentation, we’ll go over the most common sources of frustration and look at a few approaches that could be used for the container runtime to decide whether those privileged actions should be allowed or not.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Stéphane Graber</b></p><div/><p>Stéphane Graber works as the technical lead for <span class="caps">LXD</span> at Canonical Ltd. He is the upstream project leader for <span class="caps">LXC</span> and <span class="caps">LXD</span> and a frequent speaker and track leader at the various containers and other Linux related events.</p>
<p>Stéphane is also a long time contributor to the Ubuntu Linux distribution as an Ubuntu Core Developer and he currently sits on the Ubuntu Technical Board.</p>
<p>On his spare time, Stéphane helps organize a yearly security conference and contest in Montréal, Northsec, where his knowledge of Linux and network infrastructure is used to simulate the most complex of environments for the contestants.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47942017-09-04T13:42:17-07:002017-09-05T09:01:45-07:00UID/GID shifting filesystem overlays2017-09-04T20:42:17Z<div><dl><dt><b>One Line Summary:</b></dt><dd>An update on shiftfs and the next steps for remapping overlay filesystems.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>Containers using the user namespace have a different view of uid/gid than processes in the initial namespace.</p>
<p>This makes it difficult to share data between the initial namespace and other user namespaces as any unmapped uid/gid will show as -1.</p>
<p>Shiftfs is an overlay filesystem which remaps UIDs and GIDs transparently for you.</p>
<p>This presentation will go over the current state of shiftfs and the next steps needed to get it or something like it merged in the upstream Linux kernel.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>James E.J. Bottomley</b></p><div/><p>James Bottomley is a Distinguished Engineer at <span class="caps">IBM</span> Research where he<br />
works on Cloud and Container technology. He is also Linux Kernel<br />
maintainer of the <span class="caps">SCSI</span> subsystem. He has been a Director on the Board<br />
of the Linux Foundation and Chair of its Technical Advisory Board. He<br />
went to university at Cambridge for both his undergraduate and<br />
doctoral degrees after which he joined AT&T Bell labs to work on<br />
Distributed Lock Manager technology for clustering. In 2000 he helped<br />
found SteelEye Technology, a High availability company for Linux and<br />
Windows, becoming Vice President and <span class="caps">CTO</span>. He joined Novell in 2008 as<br />
a Distinguished Engineer at Novell’s <span class="caps">SUSE</span> Labs, Parallels (later Odin)<br />
in 2011 as <span class="caps">CTO</span> of Server Virtualization and <span class="caps">IBM</span> Research in 2016.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47932017-09-04T13:36:08-07:002017-09-04T13:57:36-07:00Welcome to the containers micro-conference2017-09-04T20:36:08Z<div><dl><dt><b>One Line Summary:</b></dt><dd>The usual introduction talk</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>This will introduce the containers micro-conference, the format for this year’s edition, make sure someone is taking notes and then introduce the first speaker.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Stéphane Graber</b></p><div/><p>Stéphane Graber works as the technical lead for <span class="caps">LXD</span> at Canonical Ltd. He is the upstream project leader for <span class="caps">LXC</span> and <span class="caps">LXD</span> and a frequent speaker and track leader at the various containers and other Linux related events.</p>
<p>Stéphane is also a long time contributor to the Ubuntu Linux distribution as an Ubuntu Core Developer and he currently sits on the Ubuntu Technical Board.</p>
<p>On his spare time, Stéphane helps organize a yearly security conference and contest in Montréal, Northsec, where his knowledge of Linux and network infrastructure is used to simulate the most complex of environments for the contestants.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47922017-09-04T01:22:27-07:002017-09-19T07:15:49-07:00PCI Endpoint Subsystem Status2017-09-04T08:22:27Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Status update on the PCI Endpoint Framework and next set of challenges</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The <span class="caps">PCI</span> Endpoint Subsystem can now be used to validate <span class="caps">PCI</span> Endpoint Controller in a linux environment. The talk aims to start the discussion on the following challenges:</p>
<ul>
<li style="padding-right:1em;">The <span class="caps">PCI</span> Endpoint controller can be configured to work as different Endpoint Functions. In such case what is the best way to get device ID (since getting a device id for all possible use cases might not be possible)?</li>
</ul>
<ul>
<li style="padding-right:1em;">The <span class="caps">PCI</span> Endpoint Subsystem for programming <span class="caps">NVM</span> controller</li>
</ul>
<ul>
<li style="padding-right:1em;">Overlaps with <span class="caps">NTB</span> framework and possible unification strategies</li>
</ul></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Kishon Vijay Abraham I</b></p><div/><p>Kishon is an active contributor to the Linux Kernel since 2010, developing and up-streaming linux device drivers for various TI platforms. He has worked in <span class="caps">USB</span>, <span class="caps">PCI</span> and <span class="caps">MMC</span> subsystems in Linux Kernel. He maintains the Generic <span class="caps">PHY</span> Subsystem and PCIe Endpoint Subsystem.<br />
His previous talking experience includes presenting paper on “<span class="caps">USB</span> Debugging and Profiling Techniques” in <span class="caps">ELCE</span> 2012 and “Generic <span class="caps">PHY</span> Framework: An Overview” in <span class="caps">ELCE</span> 2014, “Generic <span class="caps">PHY</span> Framework” in <span class="caps">ELCUS</span> 2015 and “overview of pci(e) subsystem” in <span class="caps">ELCE</span> 2015.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47912017-08-31T18:58:07-07:002017-09-11T18:55:17-07:00eBPF cgroup filters for data usage accounting on Android2017-09-01T01:58:07Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Describes current efforts to replace the out-of-tree Android xt_qtaguid kernel code with an upstream solution based on eBPF cgroup filters.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The xt_qtaguid kernel code is the foundation of data usage accounting and per-app network policy on Android, but it consists of ~3000 lines of out-of-tree kernel code and is a maintenance, stability and performance burden. Work is underway to replace xt_qtaguid with a mixed kernel/userspace solution based on eBPF cgroup filters.</p>
<p>This talk will cover the basic design, the progress we’ve made so far (most of the kernel infrastructure is already in place), and the challenges faced by the design, such as the limited granularity of security policies and the fact that maps are not resizable.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Lorenzo Colitti</b></p><div/><p>Lorenzo leads the Android core networking team, which maintains layer 3 and layer 4 networking on Android. He mostly focuses on userspace aspects, but has contributed kernel features and bug fixes as well.</p></li><li><p><b>Chenbo Feng</b></p><div/><p>Software Engineer on Android kernel system team. Focus on android kernel networking tools.</p>
<p>Chenbo has a Master degree from Vanderbilt University major in Computer Science. He earned a bachelor’s degree in 2014 from City University of Hong Kong major in Electronics Engineering</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47902017-08-31T11:33:08-07:002017-09-14T14:24:14-07:00Energy Aware Scheduling2017-08-31T18:33:08Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Energy Aware scheduler development</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>This talk is to provide a status update of the <span class="caps">EAS</span> (Energy Aware scheduling) development efforts in the Android common kernel, eas-dev and <span class="caps">LKML</span>. The talk will go over recent features and fixes proposed, in-flight, merged along with discussion of features in <span class="caps">EAS</span> that Android leverages such as Schedtune (which is a way to bias the scheduler to be more performance centric or viceversa than energy centric for certain tasks).</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Joel Fernandes</b></p><div/><p>Joel is a Google kernel engineer working on the Android kernel. In the past he have made contributions to the upstream kernel in areas such as <span class="caps">DMA</span>, crypto, tracing and cpufreq. At Google, he works on tracing and scheduling and is involved in the Energy Aware Scheduling efforts.</p></li><li><p><b>Juri Lelli</b></p><div/><p>Juri Lelli is a Staff Software Engineer at <span class="caps">ARM</span> Ltd. working on the Linux scheduler, with a particular focus on energy aware scheduling (<span class="caps">EAS</span>) and power management. He is also primarily involved in landing such upstream technologies on Android and ChromeOS platforms.</p>
<p>He is among the original authors of the SCHED_DEADLINE scheduling policy in Linux, and he is actively helping maintaining it.</p>
<p>He has a PhD degree from the Scuola Superiore Sant’Anna of Pisa, Italy (ReTiS Lab). His research area covered Real-Time systems, Real-Time Operating systems and Scheduling algorithms.</p></li><li><p><b>Juri Lelli</b></p><div/><p>Juri Lelli received a Bachelor’s degree in Computer Engineering at the<br />
University of Pisa (Italy) in 2006, and a Master’s degree in Computer<br />
Engineering at the University of Pisa (Italy) in 2010 with a thesis<br />
titled “Design and development of real-time scheduling mechanisms for<br />
multiprocessor systems”. He then earned a PhD degree at the Scuola<br />
Superiore Sant’Anna of Pisa, Italy (ReTiS Lab). His PhD thesis focused<br />
on reducing the gap between classical real-time theory and practical<br />
implementation of real-time scheduling algorithms on General Purpose<br />
Operating Systems, with a special focus on Linux. At the moment, he works<br />
at <span class="caps">ARM</span> Ltd., where he continues contributing to the Linux scheduler<br />
development.</p></li><li><p><b>Patrick Bellasi</b></p><div/><p>Patrick Bellasi is a Senior Software Engineer at <span class="caps">ARM</span> Ltd (Cambridge) working as a Linux scheduler specialist on Energy Aware Scheduling for <span class="caps">ARM</span> big.<span class="caps">LITTLE</span> technology.</p>
<p>He developed the <span class="caps">EAS</span> ‘SchedTune’ extension to provide per-task energy-vs-performance tuning which has been merged into the Android Common Kernel. He is also responsible for configurable tooling (<span class="caps">LISA</span>, <span class="caps">BART</span>, TRAPpy) used to analyse scheduler behaviour in product applications.</p>
<p>Previously, Patrick has been a Post-Doc at Politecnico di Milano, working in cooperation with STMicroelectronics in different projects related to SoCs energy optimization and run-time resource management of experimental many-core architectures.</p></li><li><p><b>Andres Oportus</b></p><div/><p>Works on the Android kernel</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47892017-08-30T10:55:21-07:002017-09-14T14:24:26-07:00Ion 20172017-08-30T17:55:21Z<div><dl><dt><b>One Line Summary:</b></dt><dd>What is Ion in 2017</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>One slide of status of what was accomplished last year, discussion about other pressing Ion matters</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Laura Abbott</b></p><div/><p>Laura is currently employed Red Hat as a Fedora Kernel Engineer. Her day-to-day work involves bug fixes, tending the Fedora kernel releases, and other development work for the benefit of Fedora. Her previous work involves kernel development for Android phones with a focus on the memory management area.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47882017-08-30T06:11:36-07:002017-09-09T07:13:47-07:00RT IPI schedule push/pull rework2017-08-30T13:11:36Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Reworking the IPI logic to implement the RT (and deadline) scheduler for push/pull logic</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>When at Red Hat, we found that with lots of RT tasks that can migrate, and running for lots of short intervals, it can cause an <span class="caps">IPI</span> storm on a single <span class="caps">CPU</span> having that one end up having a large latency. I have reworked the code to have a single “token” <span class="caps">IPI</span> that circles the CPUs that have overloaded RT tasks (more than one scheduled to run). I have some benchmarks that show this works. I would like to discuss the current implementation as well as any improvements we would like to make on it.</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Steven Rostedt</b></p><div/><p>Steven Rostedt has been working with the Linux kernel since 2001. He currently works for Red Hat working in their Messaging Real-time Grid (<span class="caps">MRG</span>) division. He created and maintains Ftrace, the official Linux kernel tracer, and is the current real-time kernel stable maintainer.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47872017-08-30T03:33:17-07:002017-09-08T17:19:28-07:00Scheduling under CPU capacity constraints2017-08-30T10:33:17Z<div><dl><dt><b>One Line Summary:</b></dt><dd>The potential difference in max and delivered CPU performance (capacity) is growing (thermal and peak current management), can we keep ignoring it, or should load-balancing factor the constraints in?</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p><span class="caps">CPU</span> performance constraints imposed by the Linux thermal framework, or hardware/firmware are not currently considered by the Linux task scheduler despite that many modern systems are operating under such constraints on a regular basis. The assumption all CPUs can always be fully utilized at their highest performance point (<span class="caps">OPP</span>, P-state, frequency) is no longer valid and can lead to suboptimal task placement decisions. Should we introduce static “guaranteed” <span class="caps">CPU</span> capacity in the scheduler as it already exist in <span class="caps">ACPI</span> or let the thermal framework inform the scheduler about current constraints?</p></dl><div><p><b>Speaker:</b></p><ul><li><p><b>Morten Rasmussen</b></p><div/><p>Morten Rasmussen is Principal Software Engineer at Arm Ltd. working on OS power management and task scheduling. He has been working task scheduling for asymmetric <span class="caps">CPU</span> capacity systems since the early days of big.<span class="caps">LITTLE</span>.</p></li></ul></div></div><to_s/>tag:linuxplumbersconf.org,2005:Proposal/47862017-08-29T18:54:53-07:002017-09-12T22:22:04-07:00Replacing xt_qtaguid with an upstream eBPF implementation2017-08-30T01:54:53Z<div><dl><dt><b>One Line Summary:</b></dt><dd>Describes current efforts to replace the out-of-tree Android xt_qtaguid kernel code with an upstream solution based on eBPF cgroup filters.</dd></dl><dl><dt><b>Abstract:</b></dt><dd/><p>The xt_qtaguid kernel code is the foundation of data usage accounting and per-app network policy on Android, but it consists of ~3000 lines of out-of-tree kernel code and is a maintenance, stability and performance burden. Work is underway to replace xt_qtaguid with a mixed kernel/userspace solution based on eBPF cgroup filters.</p>
<p>This talk will cover the basic design, the progress we’ve made so far (most of the kernel infrastructure is already in place), and the impact on Android userspace.</p></dl><div><p><b>Speakers:</b></p><ul><li><p><b>Lorenzo Colitti</b></p><div/><p>Lorenzo leads the Android core networking team, which maintains layer 3 and layer 4 networking on Android. He mostly focuses on userspace aspects, but has contributed kernel features and bug fixes as well.</p></li><li><p><b>Chenbo Feng</b></p><div/><p>Software Engineer on Android kernel system team. Focus on android kernel networking tools.</p>
<p>Chenbo has a Master degree from Vanderbilt University major in Computer Science. He earned a bachelor’s degree in 2014 from City University of Hong Kong major in Electronics Engineering</p></li></ul></div></div><to_s/>