 Okay, a hush came over the room. I guess that means it's time to start. I apologize in advance for the popularity So I appreciate I appreciate the fact that some of you were came even though there's no seats left And if there are any seats, please squeeze in I don't see any actually open seats So again, I apologize for that. My name is Tim bird. This is the status of embedded Linux Talk that I give on occasion And let's see if my clicker is going to work. Yep. So this is a talk that I give periodically It's not comprehensive. There's no way I can catch all of the things that are going on So it's kind of stuff. I'm interested in if if it doesn't so hopefully it overlaps somewhat with things that you might be interested in And if I if you are doing like this really neat project, and I completely missed it. I'm sorry about that So I hope to accomplish two things with this One just to let people know what's going on China the general status new things in the kernel and Open a discussion of what's next feel free during the During during this session or after the session you can get into the I think you can get into the session QA on the Virtual platform even after the talk So if you I have a question at the end of my slides if you want to answer that question From your perspective or give me feedback. I'll be on I'll try to check the chat as we go through Their topic boards on the Excel platform as well so These were the historical areas of embedded Linux focus And I think to some degree we still kind of wrestle with these today system size boot time power management and And we're you know, we'll have new SoC's with us always So we're always working on new drivers and new boards and arch support and stuff like that So the the major outline here is that I'm going to go through Linux technology The Linux kernel and then through some technology areas little bit of industry news some some comments on the community Trade associations and that type of thing and some interesting use cases this year. I've been kind of focused on space stuff for various reasons and Then some conclusions. So I don't have a lot of time. This is normally a talk that takes me about an hour and a half And nobody wants me to go that long Also, I Well, I took all this material out so but the other thing about this is I really hope that The popularity as indicated by the crowd kind of matches the the Goodness of the content, but we'll see so So oh the Linux kernel. Let's just start there So I'm going to go over versions in the last year. I'm going to pick a few items from each release That I think are relevant to embedded. So of course, there's lots of things in every release that are not relevant to embedded And Drew can I have you get me a water? Thanks So and also it's not very good coverage of Socs or or driver contributions So that's that's a bit of a shame because there's a lot of work in the embedded is there I'll try to use that when I need to And then some development stats and some company highlights. So let's look at the kernel versions. I think John Corbett said Predicting when the next kernel come out will come out is kind of like predicting when the sunrise and sunset are going to be It's boring. Nobody looks at it Because it's the same pretty much every time. So kernel is usually nine or ten weeks in terms of development cycle We usually go through either seven or eight release candidates And so this is this is what the kernels have looked like in the last year We're we actually this is a this is a rarity. We actually hit the merge window on Sunday I had to update my slides. I didn't know if we were going to go to an RC eight or or the merge window So we're in the merge window for six dot five right now. So it's a little bit too early I've been busy with other stuff this week. So I actually don't know what's in the six five kernel yet But let's just go through some of these Okay So one of the first things in five dot nineteen. So this is almost a year ago Was support for eight out out. So now we're using l format executables, but eight out out Got removed from the kernel. This is actually deprecated earlier in five dot one That's not huge news. I mean, I don't think any of us are running eight out executables anymore But it does show the stuff does cycle out of the kernel So and I'll I'll revisit that theme later the erofs Read-only file system now uses FS cache. So it gives better performance on some systems And if you were in the last session you saw that it's already a pretty smoking hot file system for read-only data So it might be even better now There's more packet drop allocations if you're using Networking and I'll I have a slide on that later. So I'll come by that You can now embed a boot config file directly into the kernel image. I'll talk a little bit about that as well And there's initial support. I don't know how many people are using loomarch CPU architecture I think it's mostly being used in China But that has been that the start of that happened with this kernel So boot config. So what is boot config? it turned out that for when you're trying to do boot time tracing you if you You need to pass a whole bunch of parameters or you may want to pass a whole bunch of parameters into the kernel And so this is a new system that's allowed that allows you to pass structured key value pairs as as just simple text file into the kernel and this can be attached to an added RD image or Embedded directly into the kernel image and so when you're building for an embedded system you want to trace some stuff going on at boot time You can pass a large number of tracing options So you you know you if you've used F trace before you can know you know it kind of can get a little chatty And so you can add your events your filters your actions your probes the fields that you want to do So it allows you to set up kernel boot time tracing more easily and if you're interested in that go read the docs that are that are upstream In Linux 6.0. Okay coming in October. We got IO u-ring support zero copy network transmission IO u-ring is kind of the new shizzle in in the Linux kernel for fast file system operations. I don't know how much people are using it embedded. It seems to be I Mean it's like when you're really really concerned about file system performance. You'll start doing some of this stuff But yeah, I mean if you are interested in that then this is something you want to look at The k-unit tests now taint the kernel. So if you're running if you're running tests and you want to some of the k-unit tests can do some kind of harsh things to the kernel and you may want to know that later on so you don't want to run k-unit tests in your like your Production environment, so you can find out if that's happened There's a new thing called the runtime verification system. So this is actually pretty important for Safety critical. It's just in its infancy this last year But I'll talk a little bit more about it later The config Android option turned out that config Android as an option did not actually mean config Android It turns so people like red hat and Ubuntu were using it to signal other things They were using some of the side effects of that config parameter just to initiate other things And so there was a long discussion and eventually they just took took it out and replaced it with something I Go read the LWN net article and you can see what the new thing is that replaces it By the way, this is this whole talk is basically a shameless ripoff of LWN net If you're not a paid subscriber go go pay him money. They do they're the best in the business I think the only in the business as far as I'm concerned with all this stuff And then the print K so if you're following real-time you may know that there's this big print K refactoring going on They're trying to do threaded print K's to make them more amenable to real-time support Developers have been trying to refactor that for a long time For the RT support, but also for other reasons print K is kind of long on the tooth. It's a little bit ugly But unfortunately The refactoring didn't make it in this time And it's still it's still not finished. So it's like anyway Back to the drawing board in six dot one BPF programs are supposed to be run in a sandbox and never be able to crash your system. Well now they can and This is actually intended that You can allow a BPF program to take the system down if it detects some kind of fault And so BPF is starting to be used for all kinds of thing Of course, it's used for network filtering and stuff like that. That's its original purpose It's been used a lot for tracing and now it's being used There's a lot of discussions about security and how BPF can be involved in that So you can do these runtime modules that get compiled on the fly Insert it into the kernel run in a sandboxed environment and and and do interesting things With with your kernel at runtime Another big piece of news on this kernel in December was experimental support for rust Okay, so if you're if you're one of those people that talks about rust at cocktail parties then Like like David then then this is very exciting to you There's improved top-level Page support and the kernel documentation just kind of continues to improve which is really good And then support for multi-generational LRU. So again, if you're a scheduler fanatic There's there's still after all these years. I've been doing Linux for almost. I gosh I think it might be 29 years now and they're still coming up with new scheduler algorithms and and page replacement algorithms, so This is a new system that kind of groups pages differently into generations and and Improves performance under some workloads. Okay. Now. This one is where it gets interesting in 6.2 The slob allocator was deprecated. So if you don't know what the slob allocator is then Bless your heart. You you missed you missed an episode of Linux history, but that's okay so this was introduced a long time ago as part of the config tiny patches and They're deprecated because nobody seems to be using it And again, I have a whole slide on this More rust infrastructure code was added there are improvements to squash FS and there's a new RV tool Which is a runtime verification? So there's a tool now upstream that allows you to manage the models that are used for runtime verification And then six dot three we're coming up to pretty recent kernels here Lots of unused arm boards were removed in this release like lots So if you have a super old arm board and you've adopted a new kernel and find out that it's missing Shame on you for not paying attention you But anyway, so stuff does leave the kernel You can't just assume that if you're not up there maintaining it helping out that it's going to stay around forever The kernel can now be configured is is geared in here Yeah, the kernel can now be configured with a built-in drystone test. Thank you for for that So that that's really interesting. I mean we're seeing a lot of more tests that are upstream in the kernel source tree K Unit and K self test But this one is like if embedded in is it in the K self-test tree or is it just somewhere else? Okay Anyway, so this is something you can run you do some performance ben trucks. You don't have to rely on bogo MIPS anymore Which is a is exactly what it sounds like it's a bogus measure of performance The make v equals zero option was removed. Okay, so There's been a trend. I think it I don't know if it's on this page or the next one No, anyway, there's a trend to make kernel much more chatty with warnings and debug messages and so I think you're not allowed to turn off the messages So if you're a kernel developer and you have and you've written some code that causes the kernel to issue the compiler to issue warnings You have to now go fix them Because Linus wants a clean build and it's taken a long time to get to the point where that's the case where the you know There's so few warnings that it's just not a bunch of noise that people ignore But that's an important attribute for upstream There was a minor change to developer certificate of origin and actually I'm going to circle back to that later So I'm going to not explain that okay in Linux 6.4 The slob at memory allocator was actually removed. Okay So there's some nice documentation that was just added for building the kernel So if you're kind of a newbie The kernel documentation continues to get filled out Oh, there's this whole kerfluffle about module license declarations so The idea was that some things Someone wanted to analyze whether or not Thing what things in the kernel could be built as modules? And so they were declaring they were using this module license if that happened to be in some of the source code They're using that as an indicator. Well, this can be a module Well, it turns out that this was used in all kinds of situations where it couldn't still be a module And so there was a lot of discussion about that it was it was had kind of been misused over the years So that's been cleaned up and then user trace events API fixes has been merged. Okay, so that's that's that's the end of my kernel thing. I did that super quick I don't know how I'm doing on time. Okay, so loom arch. I just want to talk about this a little bit I don't think very many people are using this in embedded. Maybe they are I don't know very much about it But it's a processor architecture designed by a chunk Chinese company called loom son The architecture was started in 2002. So it's not like brand new It's a variant of MIPS, but recently it kind of went through an overhaul And it incorporates ideas from both MIPS and risk 5. It's not binary compatible with either, but So it supports kind of the standard things you'd expect 32-bit reduced instruction set so that might be applicable to embedded a 32-bit standard size instruction set and 64-bit instruction set And it seems to be the interesting thing here. It seems to be a processor Architecture that was designed with Linux in mind as it's as it's kind of primary Host OS right which is different than you know a lot of other systems have been designed with other operating systems in mind Anyway, so if you're interested in that there's there's a link you can go learn about it Okay, so I got the developer stats for six dot three again shamelessly ripped off from LWN net This is kind of what people are working on And it's very characteristic. So if you see there's about every kernel release, there's about 2,000 developers that are involved And when you think about the size of the code base and the the amount of activity, that's actually not that big of a number I mean there are there are companies. Well, I my own company. I work for Sony. We have at least 2,000 developers Working on the latest kernel and not that not not all of them are working upstream So I mean there so and the interesting thing about this is the 250 new developers So every release and this is a pretty standard Stat that Every release there's about 250 people show up who have never or at least we don't have records enough to identify Who've never contributed a patch before so if you if you thought oh, well, you know You have to be you know, you have to have been working on the kernel for 10 or 15 years if you want to be a Kernel developer, that's not true at all. Okay, there's still an opportunity. You see a problem in the kernel Raise it on the mailing list go out and you can you too can be a kernel developer So this is some of the most active developers by change sets Christoph Kosolowski He's with Lenaro doing device tree update. You see a lot of Lenaro people Arn Bergman makes the list for removing code and and then you can See some of that Christoph Helwig always shows up for some reason or another well We know why he shows up because he's in the block file system. He's refactors code. He's he's like a Mega janitor in the kernel And he touches a lot of stuff This is same stats, but by lines of code changed Again, Arn Bergman with removal of old code Qualcomm added a Wi-Fi driver Greg Kraw Hartman did lots of driver or work, but he also removed some stuff from staging hands for quill removed old media drivers and Kai Hawking, I hope I'm pronounced that correctly removed lots of DRM drivers So the interesting stat on this kernel was four out of five developers Were made the top top five list Because they were removing code from the kernel. Okay, so that should tell you something about the churn that's going on upstream We are trying to keep the thing kind of sane and and maintainable In terms of commit log entries So these are not the same numbers that you saw in the previous pages because I use I have my own tools In fact, those are the tools I'm using just get log and I have my own thing called author stats But the kernel commits it's about the same every time and when you think about it It's absolutely mind-boggling that this pace of development has has been consistently like this Seriously for like five ten years 16,000 patches Going in in a seven-week period seven to nine weeks And that's I mean obviously the patches are being developed out of that window The commit window, but still and like I said consistently about 2,000 developers So here's your most active organizations by the area of the kernel they work in The core kernel you have kind of the usual suspects Google Intel Red Hat meta Oracle in architecture. No surprise. Lenaro shows up and IBM and Intel with drivers a lot of the a lot of the reasons that the drivers that like Intel AMD have Have large driver stats is because of GPU code tends to be pretty verbose File but people working on file system and block layer Networking and you'll see kind of the same people everywhere. We could use more help in documentation I know John Corbett just recently raised that on the case summit discuss list if you have If that's that's a great place to start if you want to get involved with kernel upstream development Don't know if your cc Chops are good enough Go go work on some documentation. There's always stuff there low hanging fruit and the work is very much appreciated This is this is I focused in on embedded companies I apologize if your company is not listed here and again, this is work This is contributions to the kernel. So if you see a low number there It doesn't mean that that company is slacking off. Usually it means they're up doing stuff in user space But these are some of the major companies that are and then I looked at who the top contributor was from each of those companies and The types of things they're working on and you see it. It's kind of all over the map. So you have Bailey Ray Quarantine was doing a lot of work on Rockchip and crypto libraries or the crypto subsystem Miguel Rinal was doing Network and NVM NV mem stuff and just down the list. I'm not going to read through all these But you can kind of see what's happening in terms of the the kernel and and What companies are are contributing upstream? Okay, so that was the kernel. That was my focus on the kernel Let's switch over to technology areas and I just want to go through a couple real fast So I talked about yeah fast faster than I have been So I talked about loon arch support already a bunch of arm boards have been removed There's been proposed removal for super H and itanium. I don't know if anybody in here cares about itanium But he there are I know that there are some companies that might be concerned about super H going away so if you are you need to step up and and You know get up get on the mailing list I know and embedded we're often three to five years behind the times with our kernels on our products But try to pay attention to make sure your architecture doesn't get removed So I don't think I don't think arms gonna get removed or Intel's gonna get removed But you know be just be watching the mailing list keep on top of stuff the big discussion that happens when these Architectures are considered for removal is well is that someone willing to maintain it? Is anyone you know what obviously? We're not going to try the upstream is not going to try to remove stuff if it's known that there's a lot of users But we can't tell if there's users if there's no developers if there's no activity on the mailing list and So yeah, if you if you want something to continue continue to be maintained upstream You need to be involved with the upstream, you know message forums and mailing lists and stuff like that and Seems a bootloader. Okay. This is a really neat feature So you boot now supports the ability to load images over HTTP Okay, which previously it supported the UDP protocol and you could use NFS or TFTP servers, but now it's got like I guess it's got an HTTP stack in there And actually I had this actually caused an idea. I'll have to talk to the unique people offline But you can now download your kernel from a web server So that and a web server is pretty easy to set up these days Especially like a local web server, you know, you can just have a development one on your on your host system And then snag boot which I'm going to talk about so snag boot is a set of tools that can help boot and install images on Boards that fail to boot. So this is something that bootlin produced and its open source consists of two two utilities snag recover to initialize memory and run the bootloader and then snag flash which Will communicate with the system to put a new working system image on so what this addresses is that? for a lot of development boards, they're kind of custom proprietary tools that you have to use to recover those boards and Sometimes sometimes they don't work But this this looks like a really nice system I had I have to be honest. I haven't run it myself, but but I read about it and there you go The the core kernel. Okay, so something If you if you thought they were done working on core aspects of the kernel like fork and exec You're wrong because they're even in the last little bit here. I think this was about three months ago I Think Josh Triplett did some work on I owe you ring spawn. So I owe you ring is the new way to have accelerated Operations performed in the kernel without having to go through system calls You can queue up a bunch of operations in a ring buffer That then the kernel can do without having to go through the syscall boundary multiple times and They said well if we're doing it for reads and writes how about if we did it for forks and execs and So it's actually more efficient than traditional user space based fork and exec and part of the reason for that is that Fork and exec if you know understand how they work They're discarding a lot of pages as you go through the copy on right to load the new exec image in And so you can now transfer control to a new program all inside the kernel without interaction through user space and It's 66 to 10 percent faster than v fork and 30 plus percent faster than pause it spawn. Okay, so You know, what will the kernel developers think of next? They just keep optimizing stuff in terms of file systems in I owe some work on MTD SPI nor Enhanced locking to support reads while writing. I talked about the ER of s enhancements Lots of tweaks to files existing file systems and drivers every every release. There's a bunch of tweaks where You know some like f2fs supports some new I octal that I didn't even know existed stuff like that, but Not much that I could see that was specific to embedded. Okay. Now. I'm gonna go through Languages so micro python a lot of people are using micro python and their projects. It just had a version release in April It's got a new package manager reduced code size Support for some new board support for Zephyr and support for wasm. Okay, so I don't know if that's how you're supposed to say that by the way but wasm web assembly is Allows you to take Programs and run them in a browser so you can compile down to this bytecode that runs And so you can run by thon apps in a browser now So maybe we can get rid of that Java thing Anyway, that's sorry Python 3.11.3 was released in April 3.11 in general is It's big claim to fame is it's faster. It's about 1.22 Times the speed up on a standard benchmark claims to be 10 to 60 percent faster than Python 3 and it's got some other stuff Going on so you can check that out. Oh, I thought this was interesting. So I the title of the article I stumbled across was yeah developer creates self-hildling programs that fix themselves Which is a little bit of a stretch, but not not too much. So there's a project called Wolverine that will run your Python program If it sees an exception and a trace back it will analyze the problem Using chat GP using GP GP for it will go out. It will construct a patch It will apply the patch it will rerun the program and it will continue to do this in a loop until your program has no more Exceptions. Okay, so how useful that is it's hard to say but it you know AI the AI You know I for one welcome our new AI overlords So that's that's I thought that was interesting. So we were hoping they don't do the same for Kernel programs drivers So rust support continues to go into the main line. We still do not have what I would call a real driver But rust is using being used more and more throughout the industry. In fact, there's an whole operating system Cata is and 1.71 will likely support the Musil C library And I'm gonna have to go faster. Okay, so there's a new function if you're doing networking There's a new function to provide the reason for a packet drop. It's called K free SDK reason It used to be just called K free SDK And so now in the kernel you can instrument The reason that a packet was dropped. So that's actually pretty handy gives better Diagnostics for what's going on when your network is acting up Real-time everybody wants to know about the slot of the status of real-time So sleeping spin locks were mainline in 5.16 the patches have been going in continuously But it stalled a little bit I think it actually last year it kind of hit the the low point of out-of-tree patches with about 1300 lines of code This year. It's back up to about 3100 lines of code, but that's a very small amount Okay, it's really easy to apply the the real-time patches. In fact, I have a script that does it and so Some of the changes that are outstanding are print k8250 serial driver and actually in the in the summit this morning Or not in Monday morning. I was talking to Thomas and he says those are actually kind of the same thing The reason that print k has problems with real-time is when it's talking to the serial console And so some other some other areas. We're expecting any time I think I read somewhere maybe on LWN net that maybe six dot five is when we'll get the the last patch in the series that will Allow you to enable it in a vanilla, you know kernel But we'll see I think it'll take a little bit longer to get print k refactored than people thought So so that's the status of real-time Again, it's not that hard to apply the patches and preempt RT is working really well Colonel hardening continues in the area of security Bf Bpf raises some interesting security issues when you're trying to they they want to do something called authoritative LSM hooks and that was kind of shut down that was rejected And there's been some discussion about allowing unprivileged users to run BPF modules Allow allowing an unprivileged user to inject code into the running kernel. What could go wrong? Okay, I got a comment on this I think okay I'm okay with you taking the slob memory allocator out of the upstream kernel But you didn't give it enough deprecation time in my opinion So not enough people kind of showed up on the list. They announced it six two They were gonna do it but it was only like five months between deprecation and like it's gone And so if you're using the slob allocator, I'm very sorry. It's gone now And don't move to slab because they're they have their eyes on that one But you're so there is a new option config tiny can slip config slob tiny So the new allocator you're supposed to be using is slob SLUB. Oh And this is not brand new news at all. This is actually what eight years nine years old news or something. I don't know There's a tool called bloaty McBloak face That I stumbled across by Google To analyze the size of elf binaries and it turns out it does a really really good job I mean it goes into the structures and can give you a Comparison between two binary so you change some config options and recompile you can see exactly what it does to the all the different sections the BSS data everything And so it's open source a long time ago, but I just thought I'd throw it in Oh, but in the area of tiny So there's this weird thing going on where I think it was Paul McKinney wanted to write some Some Program some test programs. They wanted to be as small as possible So they started developing this system where they're using some macros to replace the C library And this was originally created for kernel testing. It's basically just a set of dot H files Which create macros again and you it creates really really small binaries depending on what you call right because it's so I did my own little test hello world test and Hello using libc was only 11 59 bytes, but then you know it depends on like a 2 megabyte C library Hello, no libc With no dependencies on any C libraries whatsoever. It was actually smaller So I had a binary that had no dependency on C lit on libc that was smaller than the other one Though I don't know if this is useful for embedded if you're doing super super hardcore low memory embedded This might be something you want to look at it's it's not intended to you know You're not going to be able to compile chrome or something Using using kernel macros to replace libc, but for something I one of the things I wanted to test I didn't have time was to see if HTTP D light or something would would compile and that would be interesting Okay, because I'm getting low on time. I'm gonna skip the testing stuff LTP had a new test suite Tool chains keep getting released Let's see. I these slides will be upstream LLVM had another release and it's being used a lot more Oh, you can submit patches to the kernel without having to be on the mailing list It's not recommended you should be there to answer questions, but you don't actually have to submit through the mailing list That's been a that's been a stumbling block especially for people who are in companies that have exchange servers and weird VPNs and Sony mangles all the mail that goes out From from my system. So I usually have to submit patches outside of Sony's mail system. So the tools like before are really interesting There's something always going on with truth perf. Okay, Yachto project. So BMW joined the Octo project That's pretty cool. They just had a release the Mickel door release Improved memory and disk usage improved parsing time and so that's good stuff Kernel community. Okay. This was funny. I just had to throw this one in so Lina's for a couple of years now has had a set of scripts that detect when he's got certain words in his emails and And we'll prevent them from going out and so and Lina's now no longer uses those words But every once in a while when he quotes an email from someone else, they have one of those words So he has kindly asked that other developers, please do not call each other morons on the list So there you see a it's a knock-on effect from Lina's improving his behavior now now other people have to improve it as well There was a very slight change to the DCO Okay, so let me get to industry news Some of these I can go through real quick SFC sues Microsoft over GitHub co-pilot If you're using get if you using co-pilot, I would talk to you the lawyers at your company about that Well these lawyers at my company were very interested Well, let me let me talk about this. So this is similar in spirit So there's all these lawsuits now about AI and what the training data sets are and are they using it in a in a copyright? friendly or copyright compatible fashion And I think this stuff is all gonna take it's gonna take a while to go through the courts But it doesn't need to get resolved and I think the SFC my personal opinion off the record not speaking on behalf of Sony But my personal opinion is I think the SFC had a point here that you know When you're reproducing verbatim code and it's GPL code and and you're not providing, you know, the license of the source code Okay, that's a problem Intel and arm have this agreement To do designs on using Intel's 18a process By the way that a stands for angstrom. Yes, we're talking about sub nanometer processors That is Kind of mind-blowing Oh NASA and risk 5 so NASA has selected Microchip to design and produce a 12 core risk 5 so they're doing a new NASA does these types of things they designed a new processor specifically for space applications And designed for high performance, but fault tolerance in the face of radiation that type of thing But it's interesting because they may this stuff may be amenable to be used for Other areas that need fault tolerance, right? So if you harden it for radiation, you can hardly harden it for other fault tolerance situations So Leonard Leonard pottering is now working for Microsoft make of that what you will Okay, I I'm running out of time. I'm just gonna go satellites Because I've been studying space this year. So Linux is used in a lot of satellites I was I was shocked at how many it's Estimated that about 50 percent of CubeSats are running Linux as some part of the flight stack You have major constellations like Starlink and planet that are using Linux So each Starlink satellite has over 60 processors and they're all running Linux and they're using you have once you get into space It's like a whole different world. Nope pun intended But they they're using voting algorithms and fault tolerance algorithms that are really interesting redundant failover as of June 2023 last last time I counted 4,600 Starlink satellites in orbit And so you can do the math. There's a lot of Linux up in space And of course you can't talk about Linux in space without mentioning the Mars helicopter totally cool this this little little I think four kilogram Robot Helicopter has been flying on on Mars. It's performed 51 flights. It was originally only intended to do 5 But they've they now have they added support for autonomous landing site selection So there was an issue they used to be out on this flat crater floor But now they're flying up canyons and stuff and so it actually picks its landing site automatically the guys at NASA are geniuses and So there was recently concerned It they've now kind of playing hide-and-seek with it now that they're in some rougher terrain They lost radio contact because of terrain features and distance to the rover and it's very very dicey So I don't know how much longer the fact they're using a qual-com off-the-shelf Cots processor and the thing goes down to like negative 80 Celsius every night You're going to how long can this thing last? I mean if I put my phone in the freezer every night I don't think it would last two years, but it's absolutely amazing And you can see its path and there's a lot of resources you can go to look up that so Just a really just super quickly. We lost a prominent member of our community this last year Wolfgang Dank Probably most famous for his knowledge for the creation of u-boot I think we've all been touched by his code in some way or another He's a true champion of open source and he will be missed. I'm gonna skip over the conference stuff I'll talk about conferences when we get to During the closing game if you hang around that long or you can look in the slides Eland squeaky is losing its Losing its funding. That's another thing that's kind of sad Okay, I I knew I was gonna run out of time But it's just lunchtime, right? So oh Oh just a quick list of the Linux foundation projects that are doing different things in different areas of embedded So we've got Eliza Or Eliza doing safety and certification the core infrastructure Pat Project is handling Support longevity up to 10 years support for kernels automotive grade Linux drone code Oh And I don't even have time to talk about all of them. So conclusions. We're doing great If if you know of some if you know an area where we're not doing great come talk to me Okay, because I want to hear about it. I want to hear areas where we could improve and Either in the development space or in the technology space so leave me a note contact me online or send me an email and We'll try to make things even better in the future. Thanks very much and go have lunch and have fun