WEBVTT 00:00.000 --> 00:13.640 All right, so next step, we're going to have Melissa Williams going to be talking to us about 00:13.640 --> 00:15.640 K-workflow. 00:15.640 --> 00:16.640 Hi. 00:16.640 --> 00:17.640 Okay. 00:17.640 --> 00:18.640 Yeah. 00:18.640 --> 00:20.280 Good. 00:20.280 --> 00:23.280 So hi, I'm Melissa. 00:23.280 --> 00:27.120 I am a K-R-N-O-G-P-U-Driver developer at Tagalia. 00:27.120 --> 00:34.760 In today, I will give a talk, a very inclusive talk, to not let your motivation go, saving 00:34.760 --> 00:37.840 by saving time with K-workflow. 00:37.840 --> 00:43.840 So you are a K-R-N-O developer, or you want to be a K-R-N-O developer, or don't want to be a 00:43.840 --> 00:45.680 K-R-N-O developer. 00:45.680 --> 00:49.520 But at some point, you are united by a single need. 00:49.520 --> 00:53.360 Then you need a valid aid, a custom kernel, just one change. 00:53.360 --> 01:02.160 But yeah, you need to check if it's fixed something, or it's actually improving something. 01:02.160 --> 01:10.880 And then it's given change for a given distribution, or for a given device, or for a given 01:10.880 --> 01:11.880 subsystem. 01:11.880 --> 01:12.880 Oops. 01:12.880 --> 01:18.480 Maybe you can try to figure out the number of subsystems that we can have, the number 01:18.560 --> 01:21.080 of three that we can have, the kernel. 01:21.080 --> 01:27.040 So being you are not being your kernel developer, at some point you could face this kind 01:27.040 --> 01:32.920 of situation, that is, for example, I use a space developer just as you want, just want 01:32.920 --> 01:38.080 to report an issue in the kernel. 01:38.080 --> 01:43.760 And then they say, oh, there is an issue in your driver, and it's only reproducible when 01:43.760 --> 01:47.600 you are running these distribution, specific one. 01:47.600 --> 01:53.720 And then the kernel developers ask it, oh, did you check the last version of this branch 01:53.720 --> 01:58.120 to see if it's the issue is still there? 01:58.120 --> 02:02.520 And the report never compiles a custom kernel before. 02:02.520 --> 02:11.640 So they needed to read many kernel tutorials and documentation to create and build, to create 02:11.640 --> 02:15.840 a build and deploy kernel script. 02:15.840 --> 02:24.080 And finally, the report managed to check if the latest version of this branch is better 02:24.080 --> 02:25.080 or not. 02:25.080 --> 02:27.440 And then they just say, sorry for the lane. 02:27.440 --> 02:34.440 It's my first time deploying a custom kernel, and I am not sure if I did it right, so 02:34.440 --> 02:36.600 we have this variable. 02:36.600 --> 02:41.720 But the issue is a super present in this kernel branch, we can say, like, this work branch. 02:41.720 --> 02:50.600 And then the kernel developer cannot see this issue in the target device that they have. 02:50.600 --> 02:59.720 So OK, I need now to reproduce this issue and my side, but I never use this distribution 02:59.720 --> 03:02.320 because we have many distributions. 03:02.320 --> 03:11.920 So they also created almost the same script created by the reporter. 03:11.920 --> 03:16.400 The problem, you keep creating new scriptes. 03:16.400 --> 03:23.560 Every time you change distribution architecture, hardware, or even a project, even side 03:23.560 --> 03:29.080 you are company, you just change to another project and you need a different workflow. 03:29.080 --> 03:35.480 So you create another script to your new kernel development workflow. 03:35.480 --> 03:41.960 And you know you have many babies, you have a collection of my precious script. 03:41.960 --> 03:49.280 So instead of creating and accumulating scripts, save your time with keyword flow. 03:49.280 --> 03:55.560 That is more or less an example of tip-co scripts that you can have, probably, one of many 03:55.560 --> 03:57.840 of you have this kind of scripted. 03:57.840 --> 04:00.280 This is for a hardware pie for. 04:00.280 --> 04:10.360 And then this is everything that I need to create a script to memorize, to just install 04:10.360 --> 04:17.160 and deploy a kernel and not target machine, so corresponding, and things like this. 04:17.160 --> 04:22.680 And the keyword flow, I just need to remember two comments. 04:22.680 --> 04:30.080 So what is keyword flow, it's a collection of tools and software combined to optimize 04:30.080 --> 04:37.480 Linux kernel development workflow, reduce time, spend on repetitive tasks, because we do it. 04:37.480 --> 04:40.920 We are spending our life combining kernels. 04:40.920 --> 04:48.400 Standardize also best practices and ensure reliable data exchange across kernel workflow. 04:48.400 --> 04:54.440 Like you have your set up, but someone have a different one, and you are not seeing the 04:54.440 --> 04:58.560 same thing, even if you are describing the same set up. 04:58.560 --> 05:02.000 So keyword flow also have help on this. 05:02.000 --> 05:07.680 I don't know if you will get this analogy, but for me, keyword flow is kind of a power-ranger 05:07.680 --> 05:09.680 megazard of scripts. 05:09.680 --> 05:16.640 So you are just combining all your scripts to create a very powerful tool. 05:16.640 --> 05:19.320 What is the key features of keyword flow? 05:19.320 --> 05:25.400 We have many, but I think the most important one, you can view and apply custom kernel 05:25.400 --> 05:33.600 across devices and distributions, handle cross-compilation, sincerely, manage multiple architecture, 05:33.600 --> 05:41.720 settings, and target device in the same work tree, organize, kernel configuration files. 05:41.720 --> 05:48.040 It facilitates remote debugging and code inspection inspection. 05:48.040 --> 05:53.840 Standardize Linux kernel pet submissions guideline, probably you already check it on the 05:53.840 --> 05:59.440 documentation, or maybe grab a complaint that you are not following something when you 05:59.440 --> 06:03.720 just need this bug fix to this table. 06:03.720 --> 06:12.520 And then we also have an upcoming feature that is an interface where you can bookmark, 06:12.520 --> 06:20.120 apply, and reveal by bags from the Middle East, from Laurie. 06:20.120 --> 06:27.000 So this is a list of comments that we have, we can find, keyword flow. 06:27.000 --> 06:38.760 So the first subset is around to set up and configure your tool to all the different situations 06:38.760 --> 06:43.280 that you can face in your daily tasks. 06:43.280 --> 06:48.640 You have another subset that is to build and apply custom kernel, as I mentioned. 06:48.640 --> 06:57.800 We have also some tools to manage in interact with target machines remotely or locally. 06:57.800 --> 07:07.200 And to inspect and debug a kernel, to optimize best practice, to batch submission, like 07:07.200 --> 07:14.080 code style, maintainers, but also send batch, we can assure that you are sending batch 07:14.160 --> 07:25.480 to the list of recipient and mainly for this change, who needs to know about that. 07:25.480 --> 07:28.640 And then the upcoming one that is the batch app. 07:28.640 --> 07:36.000 So how you can save time with keyword flow when you are beauty in the glycosome kernel? 07:36.000 --> 07:42.840 So first, you need a config file, and probably you are like menelies, tracking and managing 07:42.840 --> 07:46.800 config files from different targets. 07:46.800 --> 07:52.320 You sometimes you are just renaming the config file with a bunch of different suffixes trying 07:52.320 --> 08:01.960 to describe or this config file is to this separate machine, and a copy of course because 08:02.040 --> 08:04.040 we love this. 08:04.040 --> 08:16.160 And on keyword flow, we can use just the kernel config manager common, and it will do, 08:16.160 --> 08:22.680 you can save, you can store different config files, describe it, and then you can get 08:22.680 --> 08:26.120 it back and apply on your kernel. 08:26.120 --> 08:31.800 And you need to build the kernel, and probably you are memorized combination of make comments 08:31.800 --> 08:39.480 and options, and you just need to run the keyword flow be, the beard from beard. 08:39.480 --> 08:47.000 And then you can build a kernel with cross compilation, C-flags, C-cash, LLVM, you can 08:47.000 --> 08:50.760 set up previously the CPU scaling factor. 08:50.760 --> 08:57.920 You can also check information about the number of modules that you have there, and you 08:57.920 --> 09:02.280 can enable compilation warnings in many other options. 09:02.280 --> 09:10.080 You also have, and then after get the config file, build the kernel, and then you need to 09:10.080 --> 09:11.080 deploy it. 09:11.080 --> 09:20.720 And then before you probably do some SSH, you copy or remove files according to your distribution 09:20.720 --> 09:28.040 and your structure, and you manually update the kernel according to the bootloader of 09:28.040 --> 09:30.320 these distributed distribution. 09:30.320 --> 09:36.360 And now you have the keyword flow D that you can do all these things like the plight 09:36.360 --> 09:40.920 kernel, prepare a target machine before to receive the kernel. 09:40.920 --> 09:46.080 You can lease the available kernel, custom kernel, or already from the distributions that 09:46.080 --> 09:47.720 is already sold there. 09:47.720 --> 09:55.440 Remove kernels, create a shareable package to our walls, and you can just put some settings 09:55.440 --> 10:01.760 like, oh, after the plight of the kernel, I just need to reboot to check if the new 10:01.760 --> 10:10.240 kusum kernel, and I need to check is okay, I don't need to SSH again, and do this. 10:10.240 --> 10:17.440 And we have the abbreviation, the combination of QR flow, the build, and the plight in one 10:17.440 --> 10:18.440 common. 10:18.440 --> 10:25.760 You can also save time, the bugging kernels locally are remotely, so previously SSH and also 10:25.840 --> 10:34.840 tap enable trace, and now you have the keyword flow the bug that you can use, you can 10:34.840 --> 10:40.760 use many, the bugging tools that we have in the kernel like events, F trace, and also 10:40.760 --> 10:49.760 the DMASC, and you can, you can handle this from your locomotion to remove target or 10:49.840 --> 10:52.680 even your own. 10:52.680 --> 10:57.840 You can also manage multiple kernel images in the same work tree. 10:57.840 --> 11:06.040 There is, we have the environment common, I mean, the environment feature, that we just 11:06.040 --> 11:19.360 kind of, you can, before you were like having multiple cloning, multiple, you are cloning, 11:19.440 --> 11:27.520 same time with poor repostors, because you need to separate, isolate your compile files by, for 11:27.520 --> 11:34.640 example, if you need to cross compile, if it's 86, and then you have an environment that 11:34.640 --> 11:42.800 you isolate both the configurations, the setup of the target machine for this, but also 11:42.880 --> 11:51.080 all the options, the compilation options, and deployment options for this specific kernel 11:51.080 --> 11:56.320 in the same work tree, so you just need to change the environment, and you can apply, for 11:56.320 --> 12:03.680 example, the same change for different setups and configurations, and also finally, we have 12:03.760 --> 12:13.280 the feature for to submit patch, we have the basic things, it's mostly a wrap from 12:13.280 --> 12:18.400 options that we are already, scripts that we already have in the kernel, but we have the 12:18.400 --> 12:24.400 sent patch that is, we automatically create the list of recipients and sent patch via 12:24.400 --> 12:33.040 mail, and also it's kind of automatized, the patch submissions rules when you need to send 12:33.120 --> 12:44.000 a mail, and we have the documentation, and it follows this rules there. So the last one is 12:44.000 --> 12:53.680 the upcoming one, it's a volunteer project, and it's a demonstration that the developer that 12:53.680 --> 13:02.960 is doing that is showing, that you have an interface to deal with Laura and then you can book 13:03.040 --> 13:10.080 mark pads, you can apply a patch or a patch series, and if you are satisfied with the result, 13:10.080 --> 13:22.880 you can reveal by and it will send automatically, revealed by for this series, then a call 13:22.880 --> 13:30.880 of action, a call to action, is stop writing, residents scripts save everybody's time is not 13:30.880 --> 13:37.520 only your time, but everybody's time, and effort with keyword flow, try with keyword flow today, 13:37.520 --> 13:46.480 and contribute to keyword flow also, because it's 100% volunteering work, nobody's paying for that, 13:46.480 --> 13:52.880 we are like a bunch of developers with many scripts that we decided to share all of our scripts 13:52.880 --> 14:00.080 and try to standardize that, and so a challenge is a place one of your scripts with keyword flow this 14:00.880 --> 14:06.720 week, and or you can make a new keyword flow feature with one of your scripts. 14:08.240 --> 14:14.720 I would prepare a demonstration, but many people just told me to not do that and live the 14:14.720 --> 14:21.120 demonstration because it will, I will get failed, so I decided to just record that. It's how we can 14:21.120 --> 14:29.200 use the keyword flow environment, so this is a list of environment, I have my local machine that is 14:29.280 --> 14:39.680 Anita one, and I have a husband by four, with the six eight six eight six six six six four 14:40.240 --> 14:49.360 husband, and I have the this thing back, so this one, I am check my local machine, and I have the 14:49.360 --> 14:57.280 Vikramas display display driver, that means we're kind of driver, and Colonel Wood's driver, 14:57.360 --> 15:02.480 and I decided to have a better description, because I was previously a maintainer of the 15:02.480 --> 15:11.200 driver, and I would like to get my baby back to me, so it's just I just put here as my precious 15:11.200 --> 15:21.360 Vikramas, and I decided, okay, I already have all the files compile, so it's just one change, 15:21.440 --> 15:29.440 so I just need to deploy, I mean, compile and deploy this change in my local machine, 15:29.840 --> 15:40.400 so I don't need to do everything again, I just need to run QWBD, and it will build and deploy, 15:40.400 --> 15:46.800 and my local machine, and then it will prepare everything in install in the Colonel, 15:47.760 --> 15:57.120 and then you can see that I didn't mention about that before, but it was like, I am using the same 15:58.640 --> 16:06.000 Colonel, so I just replaced the module there, and then I have now, when I try to get in for 16:06.000 --> 16:12.960 the Vikramas, now this is my precious Vikramas, and I am satisfied, I don't need this anymore, 16:13.200 --> 16:20.560 and I just need to, but I want to use the same work 3, that is the main line, so I will now change 16:20.560 --> 16:28.880 for a different environment using the same work 3, and now I have our 64, that is the hospital 16:28.880 --> 16:37.760 by 4, and I will use this environment right now, so what what it means, I already have all the 16:37.760 --> 16:45.520 configuration of the target machine, you can check this is the device, it can show you the 16:45.520 --> 16:56.000 information of the device, the delays because I was using my smartphone to have a connection network, 16:56.000 --> 17:06.480 so you can see it's the current, we have the information, it's the the current that came with the 17:06.480 --> 17:16.320 distribution, and you can have all the information about the device that I need to do, and then 17:16.320 --> 17:23.440 what I want to do, I will just install the plug, the Colonel, because I have all the files are 17:23.440 --> 17:28.880 very compiler, it was isolated, even if I am using the same work 3, and I just change the environment, 17:28.880 --> 17:35.520 it will just install that, and then I just said that because I would like to be a bit fast, 17:36.000 --> 17:45.200 when I reboot, and I don't need to check anything about the graphical interface, so I ask it 17:46.560 --> 17:54.080 to when I reboot, after the plug, and reboot, I will not have this graphical interface enabled, 17:54.800 --> 18:03.440 and I just pass a KW deploy and reboot the option reboot, so what will it will do, 18:03.520 --> 18:12.160 it will automatically deploy and reboot my machine, so it's sending the patch, it's a remote machine, 18:12.160 --> 18:22.080 it's not mine anymore, it's not the local one, so it will take some time, and let's see, 18:22.720 --> 18:40.800 yeah sometimes it's not that, that's why I record it, so basically it will also, I am using the upstream 18:40.880 --> 18:48.240 Colonel, and a hospital Pi has a downstream version that is more stable, but it's work, so 18:49.200 --> 18:55.360 you will see some warnings when it deploy, but it's working, it's more or less like that, 18:56.400 --> 19:05.280 so you probably will see many missing firmware, and also they will complain that they cannot find 19:05.280 --> 19:17.840 something, let's see, it's more, and that's it, so what it's say that this error you can 19:17.840 --> 19:24.880 just ignore, it will, it will work, but then they close the connection, what it means, I deploy the 19:24.880 --> 19:30.240 Colonel, and I read reboot machine, and then I try to assess, as I said, that I have 19:30.240 --> 19:35.760 read it in my keyword flow, and it's right to assess, and it's not reachable, because it's 19:36.880 --> 19:44.720 restarting, so then I will try again, when I see that it's actually completely load, 19:46.240 --> 19:53.120 and then I will see that now I have the mainline Colonel installed, you can see the first line 19:53.680 --> 20:00.880 of that, I just need to leave the machine, and then I will change again to another setup that is 20:00.880 --> 20:10.960 the steam deck, currently I am working with steam deck, so it's now I am back to the X86 setup, 20:10.960 --> 20:21.200 but what is the difference here, my laptop is in Taiwan, and this one is an indeed device, 20:21.760 --> 20:28.640 indeed GPU, and I am using a different distribution, here I have the W1, the W1, and that is the 20:28.640 --> 20:37.600 TOS that's based in ARP, and then I already have the mainline Colonel installed, but there is a difference, 20:37.600 --> 20:45.280 I try to load the VK mass, and I think it was not that motivational, as I was expecting, so I just 20:45.360 --> 20:57.040 changed one thing that when I load, when the loading of the module is completed, I want something 20:57.040 --> 21:04.320 like that keep much weightening to contribute to the carano and to do my change, so I just put 21:04.320 --> 21:13.280 this kind of message that doesn't do anything, but it's a printk like that, that we love it, I know, 21:13.360 --> 21:18.720 when we are the buggy things, and basically it's just to say let's go keep doing 21:19.760 --> 21:27.760 no meaning, it's many more traditional stuff that's like that, and then I can deploy the same thing, 21:27.760 --> 21:34.880 I will just build and deploy the carano, now it's a remote machine to, even it's 21:34.960 --> 21:54.000 86, there's a different distribution also, and what I will have, okay, it's a, I am in time, 21:54.080 --> 22:00.800 I think so, it's just a bit of delay, but it will happen, yeah, 22:05.680 --> 22:12.800 same thing in the module, because it's repair patch, it's same as you remotely, 22:13.040 --> 22:27.520 you need some time, everything's done, it's true, and then what we have here, when we deploy, 22:27.520 --> 22:33.920 we have some issues with Mkai Nitsipiaiol, but we can ignore it because we know it's, it will not affect 22:33.920 --> 22:40.640 the carano, and now I have the let's go, we came as here, and I am motivated to contribute to the carano, 22:40.720 --> 22:45.680 so basically it's that, and tech, thanks, I don't have anything, okay, 22:56.880 --> 22:59.280 questions, although at the top of the first one, yeah, 23:11.040 --> 23:16.480 hi, thank you first for creating this, I'm a kind of maintainer myself, and this looks really 23:16.480 --> 23:24.480 very promising to reduce my workload, and the first question I have is, is this also possible, 23:24.480 --> 23:30.240 this is also allow like importing patches from patchwork, and then you know, just like, 23:30.240 --> 23:35.600 basically putting them into your tool and we'll like go through the process itself and test it, 23:36.240 --> 23:41.360 and the second question is, can you also use this to automate bisecting? 23:45.200 --> 23:54.800 Okay, for the patchwork interaction, we don't have it already, we only interact with the 23:56.160 --> 24:01.840 lar right now, and it's, you know, under development, you know, it's a volunteer work, so it's 24:02.800 --> 24:10.640 not that fast as expected, and we are also, we need a contributions, but you can, for example, 24:10.640 --> 24:16.800 our open and issue, requesting this improvement, and when we have time, our spare time, 24:16.800 --> 24:26.080 start off as we, we decided to improve the, the tool, and about bisecting, we don't have yet 24:26.240 --> 24:32.800 to the same, so it's, it's a good idea, and I think if you can open and, and be sure, 24:32.800 --> 24:42.480 if you can contribute, better, or if you already have a script for that, we, we can incorporate, 24:42.480 --> 24:48.400 we can incorporate in the, in the, in the, in the, in the, in the, in the, in the, in the, just a 24:48.400 --> 24:53.600 minor falloff, I mean, it doesn't necessarily have to be patchwork, but so currently, when, when someone 24:53.600 --> 24:59.360 sends me a patch, I just look up the message ID, and I just fetch it with before. Yeah, yeah, 24:59.360 --> 25:05.440 that's something like this already work. Patch Hub, this upcoming feature, it's used before, 25:05.440 --> 25:12.720 so we have like a fancy interface, but it's actually used before to, you know, you can select the, 25:13.920 --> 25:20.000 you are not subscribed to this, you can just access this millilist and select the patch that you want, 25:20.080 --> 25:26.560 and you can just check that you want to apply this patch, and it will be applied, and also, yeah, 25:26.560 --> 25:34.080 and also, when you are satisfied with your test set, or you, you, you didn't get, get anything wrong, 25:34.080 --> 25:39.760 you can just say, okay, I accepted that, I revealed that, and we will send your revealed by automatically, 25:39.760 --> 25:47.200 so it's, it's automated already, it's under development, so it's an unstable version right now, 25:47.200 --> 25:57.120 but we are trying to finish this, complete that. I have a question. Do it support K exec, 25:57.120 --> 26:03.360 so you can switch the kernel without rebooting the machine, so be faster into testing a new feature. 26:03.360 --> 26:11.360 For module, yeah, we can do that, but we feed a built-in thing, we need to reboot, I mean, 26:11.360 --> 26:13.280 that's the way that we have. Thank you. 26:17.440 --> 26:22.480 If the module work properly, when you load and unload, because sometimes you're trying to unload 26:22.480 --> 26:24.320 the module, it is completely broken. 26:28.320 --> 26:32.080 So two questions. The first one is there any support for Android, 26:33.120 --> 26:39.680 like using ADB instead of SSH, and the second question kind of related to these rows that cannot 26:39.680 --> 26:45.360 have multiple kernels, is there any safety net in case the kernel that you just flashed is completely 26:45.360 --> 26:55.680 broken, that target is not going to boot again. Okay, yeah, not as any, you can reboot, 26:55.680 --> 27:00.560 if you have a boot loader that you know you have to grow when it's managed correctly, 27:00.560 --> 27:05.760 you can do it that it's, if you have, if you have group, but if you have, for example, 27:05.760 --> 27:12.320 you boot, right, and you have only one kernel in boot image, and then you flash it, and that basically 27:12.400 --> 27:17.760 means that you will end up in your boot that is in fast boot or something else. So it's also 27:17.760 --> 27:24.160 built for the tool to support flashing over USB instead of just going to the target and deploying 27:24.160 --> 27:28.640 through. You know, not yet. Not yet, but it's a, yeah, it's a good feature. 27:31.600 --> 27:33.360 Anything else? Yeah. 27:36.320 --> 27:41.440 Just one question about the part submission does it wrap around before, like, completely or 27:41.520 --> 27:52.000 partially or is it working progress? It's a, we already, we used before, already. It's a, I mean, 27:52.000 --> 28:01.200 many of our comments we are, we're using before features, but it's under development yet, 28:01.200 --> 28:08.720 it's not completely. Hi, thanks for the talk, a tiny question. Because I'll see that there are some 28:08.800 --> 28:14.080 system-debindings. I wonder if there is support for other stuff such as Rani Tore OpenRC. 28:15.600 --> 28:22.240 If there is other, no, it's not yet. Yeah. I think we'll be right up there. Thanks a lot. 28:25.680 --> 28:33.360 The last, I don't know, it's the last thing is, I see you have many ideas, probably you have 28:33.360 --> 28:40.080 many scripts. So if you already have it in your machine, maybe you should consider to just 28:40.080 --> 28:46.160 you know submit proposal to us. If you are scripting, we will be, make our life easier.