 Okey, mari kita mulakan. Okey, saya pergi ke konferensi Linnos Embedah di Frans. Ya, dia bukan Julu, saya tanya Rondi. Okey, ini Oktober. Maaf, saya tanya, sekarang saya faham. Okey, ya, jadi konferensi ini bukan hanya konferensi Linnos. Ini sebuah konferensi dengan emfasis pada sistem Embedah. Sehingga nama adalah Linnos Konferensi Embedah. Okey, ini berikutnya. Ia mempunyai treks dan terakhir selama 3 hari. Dan saya cuba berbincang dalam lebih kurang 20 minit untuk anda. Okey, saya akan cuba. Jadi saya hanya membuat beberapa tanda-tanda tanda-tanda. Jadi tanda-tanda tanda-tanda pertama diberikan oleh Linnos Konferensi Embedah dan Fella dari Linnos Embedah. Dia beri tanda-tanda tanda-tanda daripada sebuah tanda-tanda CPU, seperti Melodon Spectre Fowl dan Zobey Land. Jadi tanda-tanda ini adalah tanda-tanda tanda-tanda tanda-tanda. Saya ingat bahawa apabila pakaian CPU pertama diberikan, Microsoft, Apple dan Linnos Manus akan pertama dikotong. Tetapi WN tidak dikotong sampai 48 jam sebelum pakaian ini. Dan sebabnya WN tidak dikotong. Selain sebab 80% Linnos Insolasi, tanpa Android, mereka menggunakan pakaian WN. Mereka sangat sedih. Mereka mengusahkan bahawa bahawa wadah ini lebih berlainan lebih daripada 2 tahun lalu, mereka masih mengubahkan pakaian. Mereka mengubahkan pakaian securiti dan masih mengubahkan kemampuan. Mereka mengusahkan bahawa kita harus selalu mengubahkan perak dan bios. Ya, ia berhasil. Bukan hanya mengubahkan pakaian securiti dengan CVE. Kerana tidak semuanya ada CVE. Dan pakaian normal tidak akan dilakukan sebagai pakaian securiti 2 tahun lalu. Tapi sekarang mungkin tidak ada masalah sekurang-kurangnya sehingga selanjutnya Jadi dia berkata, update semuanya Dia kata, tidak saya kata Okey, so another talk that I attended was called NuttX for Embedded Linux Developers So NuttX is a POSIX compliant real-time OS So you can think of it as something like 3R TOS So it supports many microcontrollers, small footprint and runtime memory Okey, so the key thing for this OS is that it's POSIX compliant It aims to make app development easy for Linux developers So you can basically write an app and test on Linux Before porting it to a microcontroller So this guy is a speaker from Sony And he said that Sony has used NuttX in many of their audio products Like wireless headphones, recorders and music players Okey, so something to consider Okey, so next talk, safety vs security This guy he basically try to give us a viewpoint from both sides of the table Okey, so what's the difference between these two cultures So let's look at safety So safety people will say that code must be proven and certified And the bug lalinos goes down with time The longer you don't see a bug, the less likely there's a bug in the system And a known bug with no consequences should be ignored If your bug doesn't cause you any problem, don't care about it And upgrade only as a last resort Because upgrading something has the potential of breaking something But the security people will say that You must react quickly no more to discover a security bug And track models will evolve at that So all bugs are potential exploit and you must fix it quickly And therefore you must always use the latest version So in conclusion, so any change is a risk and needs to be justified The safety people say The security people say any bugs are potential security weakness and must be fixed Okey, so again there's these two clashes already Okey, so that is it for the high level view How does this apply to the embedded world Okey, so the safety people will say Okey, if you want to do an upgrade Your upgrade must be robust And you must deal with failures on your own Because you have all these issues It don't have access to your product It's not like a server where you can easily access it Right? Your product is somewhere far away, maybe And let's say that I block something wrong with your file system Or your disk Right? You have conflicting configuration files If valid user configuration All these that may prevent your system from booting Properly And not to mention your kernel sometimes has to be upgraded And some systems can't be stopped It's not like you can just reboot your server anytime Right? Some mission critical stuff You don't want you reboot It may lose some functionality Your old hardware can't be face out easily Once deployed there It takes a lot of resources to switch the hardware out And the last point is that deployment time is controlled by the user You cannot just anytime say I want to stop and upgrade It's determined by whether your user say You can turn off the system or not Okey, for the security people This is how they will think Your physical access you can't be restricted It's not like your data center You can restrict via lock and key Certain things like your bootloader attacks Your JATAC attacks All this They can affect your system If your attacker has access to the hardware And if thing is There's no upgrade culture Because you need to see where embedded systems are being deployed Who are the people that develop this? So for example, hardware makers The guy give example Of acorns So when you Design acorn You sell it to customer You just leave it You don't really need to update it But the moment you put the Embedded computer inside Then you need to take care of updates And these people may not have the mentality To do it correctly Or People who work in start-ups Who May not have a long term maintenance team Because the start-up may not survive long enough For that Okey, no offense to anybody who work start-ups here Okey Okey, so what is the solution? So one way is you Avoid the problem Or put it nicely Don't care about it Okey, why? Because this point is Not all products are safety critical So if your products are Not connected to internet And you can guarantee that no one else can access it Then it's fine really Right But you still need a robust upgrade system In case you find really genuine bugs In it You may need to accelerate re-certification Like you need to make sure That if you want to do an update There's a quick way to certify all your upgrades Basically safe They don't break your product You can try to isolate certain things Like containers, hypervisors The hardware separation one So maybe you can Like for example separate out the mission critical Part into a separate microcontroller So while your main system is to be upgrade The microcontroller can run the mission critical Function separately And just have a culture Okey To make sure you plan an end of life for your products So this one a lot of people They To not really think about it so much Like make sure You start off with an end in mind Okey This is another talk to Reduce the boot time reduction techniques Okey So this guy is trying to Reduce the boot time of his big bone black Okey So he has this function He has an LCD, a webcam And a microSD card Basically to record something And to display it Okey So he say How fast can he reduce the boot time So his initial boot time was 9.45 seconds Okey So this was his principles Okey You start by optimizing things That won't reduce your ability for future measurement So you may need to instrument The measurement of a boot time And if you don't want to Basically go and optimize it Such that until you cannot even measure this You know make sure that you keep your measurement tools Until the very last minute Okey Another one is to try to keep to slower storage So the example that he gave in his talk was to Deliberately use a very slow SD card So it amplifies the boot time So he can easily see that Okey, certain things take this much longer to start If you use a very fast SD card You can't really see a difference Okey And start from last part of boot time Means you finish by the kernel at the end By the boot loader You start from the apps that you start And work backwards Okey So I will skip right to the end of his talk And just show his summary slide Ya So this is You can see that the biggest one Is the silent kernel Basically disable all the messages The tracing one This one is actually the instrumentation So you can see that the instrumentation itself Will take up About 550 milliseconds So it takes a lot of time But you should do it at the last Okey So you can take a look at all the others That Automization that he has Okey, you have to know more Just go and look for his talk Okey, so Linu Sofars Ya, finally see this famous guy Okey No need to intro who you see Okey, so I like this exchange Okey, so the other guy Is an open source officer at VMware asking What do you do in your job as kernel maintainer Oh, then you say Linu say I read lots of email and sometimes say no You need someone with a background to say no Occasionally I will send out slow code As suggestion But I don't write code anymore So Dirk said I think we just de-grorified your job Ya So he is more of a managerial role now It doesn't write any code now already Okey Okey But of course this exchange is lasted for half an hour Ya, so you can watch the videos all online Okey So this one another speaker Also about boot time optimization But closer to the system D-side Rather than the lower level kernel aspects Okey So just a quick intro like in it So in it is basically something that Start after the kernel has started And it starts all your stuff Okey, and configures everything So system D is in a way Something like a replacement But it's not just an in it He adversize System D is a way of life It takes the work freaking everything Okey, so you can see This are the The stuff that he has Ya It takes charge off Okey So Okey, so this slide This talk is very long But the idea is She was trying to point out is that System D although it is used for servers You cannot just take a configuration That is optimized for server And then dump it to your embedded system And hope that it works Hope that it works the same way Okey So Okey So what the generic principle is Still the same You leave out task you don't need And you change the order of task Of course the high level This high level principle To actually execute it You need to benchmark your system Right So you benchmark With something called system de-analyse Which actually I did not know about this Until I tender his talk Ya So you can take a look So you just write this command Then you can see roughly Like each service takes how long to start Then you can optimize accordingly Okey Ya as usual You want to know more Just see his talk Ah, this one I found was quite funny Linus Ice-Quessy in the 21st century So Okey So friend sang right He is a freelance consultant But he consults Mostly for renaissance Or chip maker And on his sideline He is an Ice-Quessy maintainer For Linux So ah He first started off with What can go wrong with Ice-Quessy Right Okey so He puts it very simply Like Okey The basketball has not changed since 1982 It's not rocket science So the drivers are relatively simple Right Okey But he points out these three points Ah Okey It's a fly by subsystem Okey so why did he say that It's because people mostly don't care about it Until it impacts them Okey And mostly used by embedded purposes only So the number of people who are concerned With Ice-Quessy is further narrowed And the most dangerous thing That is largely maintain is better On one maintainer Which is himself Okey so this was the point that he made So one day he said He was quite sick ah So he emailed to a few people He said I've been tied to a bed for a few days Not in a condition to really work on Ice-Quessy So I think it's annoying sure but no catastrophe However it shows that I'm the single point of failure For Ice-Quessy patches What I don't like So like I said before I'm open to group maintenance sheet If you think you are a reliable candidate Please get in touch with me Okey ah So he sent to a few people And only one guy reply And guess what the reply is Okey I show you ah Get well soon That's right Wah Wah this one I think you have to point out That this Ice-Quessy maintainer Is kernel space Not user space So it's not the ones that you use For your hobby hobby Okey Those are user space Ice-Quessy providers Ah okey This one is kernel space Which is a lot harder A lot harder Okey Okey Okey Need to emphasise ah Ya ya true true Need to emphasise Where they are right Oh ya Linus kernel Ice-Quessy maintainer Okey Let me Okey so this was the point that He will raise that was Quite a new thing and complex Because Ice-Quessy is not just It's not static ah So this was the example he gave Ice-Quessy encapsulation With multiple gigabit Multimeter zero links Very long ah So the point was saying that Some new chips being developed Wah they Have this They Ice-Quessy okey Ice-Quessy normally is meant for Very close communication right Between electronic components right But people have used Ice-Quessy For this they go and encapsulate it with some GMSL links which normally use in coaxial cables And it will pipe to some far away Camera sensors for example Is Different lah For your typical use case of Ice-Quessy And the thing is all these are the same Ice-Quessy addresses So it becomes very hard to communicate And then the So someone I think it was the I think it was the maker of the chip Proposed to have a dynamic Ice-Quessy address Allocation Something like your DHCP server So suddenly This one is acting like Okay I need to hand out Ice-Quessy addresses So it's very Strange Okay so he So the point that he will say He's trying to make is that Clearly cases where Linux is not prepared for this This was his exact words On his slides Okay so he detailed some API changes I don't really use them But the key point is Some breaking changes Right he purposely changed The parameters to breaks People's code during compilation Yeah To force you to Read about them Because if you just change the API without changing the parameters Your code compiles fine right Or you may face problems Your code will suddenly You compile properly but suddenly fail Then another API that he also Mention was Something like a dummy device Okay so He says that Now the current drivers I think you can only I look use one address But some Ice-Quessy devices They may use two addresses So the current hack Is to use An API called new dummy device But when if your driver uses this Right Nobody else knows who Reserves this Ice-Quessy address So he proposes Proposes that People use the new API That let you supply the name parameter So they know which driver uses it Okay I left 4.5 minutes Okay Good That's in time So I have a I Join this conference closing game Yeah so This one is quite Unique Okay so basically At the end of the conference Everyone will stand up And then if you answer The conference quiz correctly You sit down Then if the last few people Who remain standing will get prizes Okay so You can see this one Of these questions That You can see Yeah okay Okay So the first stop program machine Was a loom that used sponge wooden cards Automatically with fabric designs This machine was invented in France So a green is true or red is false So which one do you think it is Yeah it's true ah It's true Okay so you can see the green and red cards right So you basically erase all Yeah So the A unique point that I found interesting was They say at the start right When they first started this conference series Both cards were of the same size So they had a problem with colour blind people Yeah Because they could not differentiate So now they make the green card bigger than the red Okay Okay so the Yeah Then This is an interesting question Which continent has the highest percentage of linux users Okay asia is what Percentage ah It means percentage to population Yeah percentage one Yes that's asking the right question Yes Because it will depend on the population or not Yes So it's a share of the continent Percentage Oh it's percentage of the population of the All linux users are a population of Of the continent Of the continent Yeah la Of the continent Yeah of the continent Of the population Neh Highest percentage Is it Europe or not? Europe or not? Europe or not? Europe or not? continent Yeah so far all wrong Because android is the next No Okay let's discount android from this Throughout the whole Throughout the entire conference right They keep minusing android Yes Right Yes So the percentage rafi 12% Because a lot of sighties are working there Very trick question right So you see There is green and red now the above Wah this this question killed a lot of people Trick question Okay So i can't come to the closing talks ah Closing talks of this So There were 10 tracks So even i personally couldn't even attend Majority of the talks la I already attended will be 10% of the talks So you can take a look at the videos Of the conference I'm not sure if they have been made public Let me try ah Sorry Ah sorry Ah Sorry Is it playlist? Oh shit Oh sorry Okay i'm not sure whether this Ah This playlist is public yet Because it was originally sent to all those who attended the conference But it was all They are all here la You can take a look how many ah 88 Yeah a lot of it Okay so you can just take a look if I'll post it on headwear later Then ah I feel They are all in English Yeah But everything is in English here Yeah Okay so ah The interaction with people So i find it quite useful to Attend conferences Because you can Talk to people and then you get recommendations For example the nutx talk right Was because someone recommended to me And i knew about assistance of open source Consultants because someone recommended to me Okay i didn't know about open source consultants before Like people they help you to upstream your drivers for example All right Okay So that At 10 seconds so that's all Just in time Okay thank you