Untitled Linux Show 159 Transcript
Please be advised this transcript is AI-generated and may not be word for word. Time codes refer to the approximate times in the ad-supported version of the show.
00:00 - Jonathan Bennett (Host)
Hey, this week we're talking about swappiness, we're talking about Proton of both varieties, talking about Wine, Nexus mods coming to Linux, KDE, Plasma news, lots of kernel news and something real interesting may be coming to Fedora 42. You don't want to miss it, so stay tuned. Podcasts you love.
00:21 - David Ruggles (Co-host)
From people you trust.
00:23 - Jonathan Bennett (Host)
This is Twit you love from people you trust. This is the untitled linux show, episode 159, recorded july the 6th. It's called a movement. Hey folks, it is saturday and you know what that means, and it's time for the untitled linux show where all things geeky, all things Linux happen. And we've got quite the show for you today, quite a few things to talk about. We've got a crew of guys and can you believe it they were teasing me before the show that one of these things is not like the other. Can you guess what that is? Send us your. If you're watching on the social network formerly known as Twitter, send us your guesses in response to this post. If you're on YouTube, give us a like and a comment. Do all the YouTube things right.
01:12 - Rob Campbell (Co-host)
Well, jonathan, if you really want to get picked on, you might want to take a look at the date you have on the show on the screen here.
01:19 - Jonathan Bennett (Host)
I am not the one that put the date there. Not my job, not my fault. We can go ahead and fix that.
01:29 - David Ruggles (Co-host)
look it's fixed, I fixed it uh, these guys all comedians.
01:32 - Jonathan Bennett (Host)
So rob is going to start us off with the actual stories, and uh, buckle up, because I have a feeling it's going to be more of the same. Uh, rob, what's, what's your slightly silly story that you're going to start us with?
01:47 - Rob Campbell (Co-host)
So last month we quickly squeezed in the news about how Linux passed 2% threshold on the Steam survey. For some time now, linux hovered in the 1% range and we were really hoping to get past that into that 2% range, make that 2% the new 1%. So you know, every time we would hit a milestone one month, we would go back down and down down the following month and be below that milestone, but still the overall trend continued to keep going up and up overall. So well, once again we have dropped from that all-time high of 2.32% that we hit back in May, but we've stayed above that 2% threshold still at 2.08%, and well above mac OS is 1.31%, which I think we just passed less than a year ago. So we've been maintaining our lead against Mac OS.
02:53
As far as gaming goes, make sure to vote on the steam survey on Linux Any chance you get. Keep us above that 2% threshold and working toward the 3% threshold that's. That's a new goal. And with things like the success of the Steam Deck and what they've been doing for gaming and Microsoft alienating its concerned users with their AI, recall, spyware, and then just the overall quality of Linux these days, I think we have a real chance at major growth. I I've even had several friends, co and co-workers actually ask me about switching to linux due to the microsoft recall thing, and these are from people that I never really expected these questions from. So you know, I I'm I'm seeing a lot of talk about this, even though they've kind of backed down a little bit on the defaults. At least, I've seen a lot of talk. So I'm really hoping this is the year.
03:57
You know, and also in other stats I found just to add on top of this when at statscountercom their stats on desktop operating systems, linux takes a 4.04% of the market share and Chrome OS, which is really just Linux, at least for now If you heard another story we did a few weeks ago. Maybe it won't always be, but for now it is also just Linux. So Chrome OS adds another 1.94%, making that total Linux desktop share around 6% market share. You know there's some Android desktops. Probably maybe they don't make a percent, but they might give a little fraction to that, make a percent, but yeah, they might give a little fraction to that. So, with all that being said, this we may just be looking at the year of the linux desktop it's a little late in the year.
04:56
Yeah you're supposed to say that in january I'm gonna start on day one. I mean, come on, what do expect?
05:03 - Jonathan Bennett (Host)
How many times have people predicted the year of the Linux desktop for one reason or another? Well, this is my first time, so Is it? It seems like we've done this long a dance before. But hey, we've hit 2%, so I mean that's a win, yeah.
05:23 - Jeff Massie (Co-host)
I'll say keep in mind. They do not publish their overall user count. So if they doubled the user count and we only went up by three quarters, it'll show a loss, even though we gained a tremendous amount of linux gamers indeed, indeed, um, you know it, it is interesting the the number of people that saw what microsoft was doing with windows and started thinking about jumping ship.
05:53 - Jonathan Bennett (Host)
I've seen, I've seen people that I follow like for their art or for their web comic, and suddenly they're they have posts like by the way, this comic was today produced entirely on Krita under Pop OS or whatever it's like. Wow, that's surprising, but there's a bit of a momentum gathering and that's fun to see. It is, it's neat, called a movement. It's called a movement. I think we just got the show title too. All right, I am going to take it away and I'm going to warn you all about something fairly dire, sort of it's not actually that bad, but you kind of have to treat it that bad Regression and this is the vulnerability in SSH and it's kind of a big deal, particularly and I will tell you, I will front load this with here are the people that really need to worry, and that is if you are running, I think, it's latest Debian is it Debian 12 or Debian 11? It's one of those two, I think it's Debian 12 or one of the RHEL 9 derivatives and you have SSH exposed to the internet, particularly if you're running a 32-bit install of one of those two, you really really need to go and update your OpenSSH install and then either restart the SSHD daemon or reboot your machine, because there is a flaw in OpenSSH that allows, after about five to six hours of constant hammering on, an SSH connection will allow someone to run arbitrary code, in this case as root, on your server. It's pretty bad. So we're going to talk for a little bit about what's going on here, because it's really fascinating. We need to learn something about the way Linux works.
07:50
Unix, indeed, works on the inside and it starts with signals. So signals is when your operating system tells a program something changed. So, for example, the SIG HUP one of our users, one of our listeners, at the end of every episode says SIG HUP as his way of saying goodbye, and that is the signal hang up, and it means in the old days that the serial connection that a terminal was connected to has disconnected, the user logged off. Well, these days it means that you have closed your terminal or you've logged off or what have you, and so the program knows to end gracefully. One of those other signals is called SIG alarm, and that happens, that gets sent when the process sets a timer. So the process says to the operating system hey, you know, remind me in 900 seconds, remind me. And so okay, os says okay and starts this timer going. When the timer goes off, it sends a signal to the process.
08:54
The way signals work is because the operating system does the scheduling. Operating system controls when the program runs. When it doesn't, the operating system just stops, just pauses like hits the big pause button, and execution in the program code itself just stops immediately. And this is the weird part what the operating system does, it doesn't do anything else, it just leaves execution there kind of hung. And then the operating system calls the function that has been selected, as it has been described as the signal handler. And so you've got this really weird case where program execution is actually entering the program a second time, and so it's a known problem that that can be problematic. So it's a known problem that that can be problematic, and so you write your signal handlers. To be very simple and to not do very much, one of the approaches to this is inside your signal handler you set a flag in memory and you immediately exit, and then your program loop continues and it finds the flag and knows that the signal has been raised. It knows that the signal has been raised. It is a known problem that not all functions and here's the term are reentrant, and that means it's not necessarily safe to enter a function again. So the function in question for the SSH flaw is actually malloc memory allocate, and what the researchers discovered was that it was possible to. So let me start here.
10:35
When you make an SSH connection, there is a thing called the grace period and that is the amount of time that can go by before you are fully logged in. So you know you go to SSH into something and you're going to be sitting there at the password prompt. You know, assuming the SSH install accepts passwords, well, if you just leave the machine, if you just leave the connection there, it will eventually time out. That, by default, is 900 seconds. The way that timeout is handled is the SIG alarm. It is a signal handler. One of the other things that you can do with an SSH connection is you can send public key material to be able to send your public key to do a key login rather than a password login, and it is possible to send that all the way up to the end of that grace period. So what they discovered that they could do is they could sort of trickle that key material along and send the last packet of key material at essentially the same time that the timer goes off goes off.
11:50
And what they're looking to do in this exploit is the SSH daemon is allocating memory to be able to store the key material that's sent in over the network. It allocates memory, it's in the middle of that memory allocation and then the signal goes off. Well, one of the things that this is where the actual flaw is, this is the actual vulnerability is inside the signal handler. In the versions of SSH that are vulnerable, it writes output to the log In the signal handler. Well, writing to the log is not one of those safe functions. See the term they use is async signal safe? Again, we're talking about asynchronous. It's not exactly asynchronous because we're pausing, but still it's the term they use asynchronous and then running on the signal. In this case it's not safe to write to the system log. One of the things that happens when you write to the system log is you allocate memory to hold the message that's going to get sent to the system log and so you have this hung allocation. That's partway through allocating, and then you allocate more memory and so you end up having this case where you have both of these pieces of memory allocated on top of each other and so you have memory corruption, you have memory confusion and so you have a race condition, basically to be able to make both of these things happen at the same time so you can get into this state. And then they figured out and there's a lot more detail in the link in the show notes. You hit that link and then you get to the link to the actual research. They figured out that they could just put shellcode in their fake authentication key and then there was a virtual pointer that they could overwrite and just jump execution into that key On 32-bit.
13:34
So one of the other things that really caused a problem is on the 32-bit installs, the address layout randomization. Aslr is address space layout randomization is entirely broken. So what ASLR is supposed to do is randomize where in memory things are at, and glibc on a 32-bit install can be in one of two different locations. So ASLR, the randomization there is all of a single bit of randomization. So they say their math says that you can pull this attack off in two to three hours, but you have a 50% chance of guessing the right location, and so they said the whole thing is going to be four to six hours. On 64-bit installs, there's much more that. The ASLR is still somewhat broken, but it is working a lot better, and so it is non-trivial to pull this off on a 64-bit install.
14:29
The versions of OpenSSH that are vulnerable is version. Let's see, I've got it here. 8.5 P1 is where the regression was introduced, and then it was fixed in 9.8 P1. So if you're anywhere between those, you need to upgrade and make sure you've got either the updated version or a back-ported patch. And one other thing to mention about this. You know we talk about things like fail to ban. Generally speaking, fail to ban is going to keep you safe from this, unless you end up in the case where you're being targeted by a whole bunch of different ip addresses. But I don't think that's really the sort of attack that is going to be popular with this. So fail to ban really makes this a lot harder to exploit. But that is it, and make sure you're up to date. Boy, it's a, it's it's. It's not as bad as it could be, but it's pretty bad so my you were saying it was a relatively recent change.
15:30 - David Ruggles (Co-host)
So my old centos 5 box it's open to the internet, still safe, right no, you should not have a centos 5 box that's open to the internet.
15:41 - Jonathan Bennett (Host)
It is interesting that I say it's a recent change, because it is a recent change that reintroduced an old problem. Right, so it. This is something that got fixed in like one of the five dot releases of open ssh.
15:57 - Rob Campbell (Co-host)
Um, that's why it's regression yes, it's, it's literally a regression.
16:02 - Jonathan Bennett (Host)
Yes, let's see if I can find so. 4.4 P1 is where they fixed it the first time, and then an 8.5 P1, they did a code refactor and the problem got reintroduced, and then 9.8 P1 is where it got fixed. Now the interesting tidbit. That is pretty fascinating. Can anybody tell me?
16:27 - David Ruggles (Co-host)
what the P1 means, patch 1, I don't know.
16:30 - Jonathan Bennett (Host)
Portable. It refers to it being portable. Openssh is maintained by OpenBSD. Openbsd is not subject to this because OpenBSD actually wrote a log handler function that is reentrant, that is async signal safe, and this is actually how this bug snuck in. But because there's so much of a difference between OpenBSD and Linux, they have a second like it's not quite a fork, but it's a second version of OpenSSH that runs on Linux and it's their portable or their P1 release.
17:08 - David Ruggles (Co-host)
Okay, an ssh that runs on linux and it's their portable or their p1 release. Okay, so a little bit. So my first question was obviously uh, tongue-in-cheek there, but a little bit more serious question about signals. Um, are these signals the same sort of signals that you can send with kill? Yes, okay I was thinking they were, but I wanted to be sure I wasn't off base, yeah it's a.
17:24 - Jonathan Bennett (Host)
It's a really really simple form of interprocess communication and it also goes like these signals. They go all the way back to the original bell unix, like way back in the very beginning it was. It was one of the really really early ideas that you know makes unix so cool. It's just maybe one of the reasons why it's done in such an odd way, with completely freezing execution and all that all right. Well, let's move on then, if the peanut gallery here doesn't have anything else to add. And, david, let's talk about wine. Merlot, the white wine, rose, no, no, no it is.
18:03 - David Ruggles (Co-host)
Wine is not an emulator. Self-referencing.
18:13 - Jonathan Bennett (Host)
Acronym those are fun.
18:15 - David Ruggles (Co-host)
Yes, always. But wine 9.12 is released Not a whole lot. You know the normal updates and everything. A couple of interesting things that I saw when I read this. First is they have initial support for user 32 data structures in shared memory. That should continue to expand their support of Windows applications that may need access to that. And then they've rewritten the cmdexe engine.
18:56
And you don't, I mean Windows is all about point and click and stuff, and so I mean those of us with linux and everything, we're more comfortable in the terminal anyway, typically um. So you don't talk about cmd a whole lot, but cmd is, I think, since windows 95 um is the terminal process under windows um, and so they've rewritten the engine that supports that. And then they've fixed a bunch of various bugs. They've added um fixes for specific games because, as we've mentioned before, wine is tied very closely to the proton um project and so a lot of protons stuff gets backported or upstreamed into wine as they release that. So it it's just a bunch of progress is being made and Wine, which is not an emulator, continues to advance the support of Windows applications on Linux.
19:54 - Jonathan Bennett (Host)
Yeah, one of the fun things with this release is apparently Firefox 126 was not working and they got that working, which that's the I think one of the latest. It was released in May of 2014. It's recent, just one version back not working and they got that working, which that's the. I think the one of the latest was released in may, but it's recent, just just one version back.
20:09 - David Ruggles (Co-host)
That's a critical fix yeah, I mean, you can't you can't run firefox on linux after all?
20:17 - Jonathan Bennett (Host)
yeah, why would you?
20:19 - Rob Campbell (Co-host)
run firefox under wine, since you can run it natively because they don't't know how to change the headers to make it look like you're coming from a Windows machine.
20:31 - Jonathan Bennett (Host)
Or you've got some extension that only run under Windows, or you've got something else that needs to talk. So, like games or what have you, their installers will often try to open a browser, and trying to use the built-in browser is not necessarily fun, although I think isn't the. I want to say there's actually a built-in browser on wine that is based on firefox, the gecko browser. Oh, that's I'm. I'm pulling some old memories up from that, though, anyway, I think so, but it's been quite a while.
21:06
It's been a long time since we had to care about it. Right, it's just worked. Yeah, it's kind of nice.
21:11 - Jeff Massie (Co-host)
And I'll say that the CMDexe, that's good because you know there's still games that when you're first installing them in Steam you'll see like a DOS window open up and you know installing and doing a bunch of stuff in the background, and so it's. It's still actually used once in a while for at least installs and whatnot. I don't know once the game is actually going I don't think so but pulling in various libraries and whatnot yeah yeah, good to see it it's.
21:40 - Jonathan Bennett (Host)
It's always fun as well to see the different things happening between wine and proton. And then you've also got wine staging and then you've got things like the glorious egg roll mashup between those, and so some things will land in wine first, some things will land in proton first, a lot of things land in wine staging first, and it's kind of this process of weaving to get all of those updated in all various places.
22:07 - Jeff Massie (Co-host)
It's fun, it's fun to watch, well, and why? Wine official is a little bit like the kernel 2 where they really want very well structured uh code in there and it it's got to have a lot of portability and things like that. So that's one reason to some of the development. The protons that you know might not work on all platforms but they throw it in that development where it's okay. They'll take a little uh right with with wine proper.
22:35 - Jonathan Bennett (Host)
They really do not like game specific fixes. They want fixes that make sense for everything and they they really shy away from from putting code in there. It's for like hacks for a specific game. But over Proton they don't care, they do that. That's their thing. That's part of the reason why. All right, jeff, do we want to talk about something fun coming to 6.11?
22:59 - Jeff Massie (Co-host)
Yes, all right. So while we're on 6.10, in the RC phase release candidate AMD is busy working on what they'll want to put into 6.11 when the pull request opens up, which is estimated about two weeks, you know, assuming nothing goes wrong, you know, and with the current release cycle, which is looking pretty normal for the RC6 release, you know, according to Linus anyway. So we'll probably we'll probably see 6.10 become finalized pretty quick here. Now this code confirms that there's new hardware on the way. So the Ryzen 9000 CPUs are going to have some integrated graphics which are based on the RDNA 2 engine. Along with this, we now know the minor versions of things like DCN, gfx, vcn, sdma and MPO IP blocks, and I'll let the listener take a look at the link in the show notes if they wish to see the specific versions, as me reading them all out I think would just be a bit boring as I rattle off a whole bunch of numbers. All these code blocks also have versions for the StrixPoint AI300 chips and they're going to be included, including the RDNA 3.5 graphics on the StrixPoint. So the other big things coming with this code is the plumbing for the ISP hardware block.
24:20
Isp stands for image signal processing. It's a hardware block that'll be found in some of the mobile CPUs for things like laptops. There's a thought that this wouldn't be seen until the RDNA 4 graphics, but it looks like it might be showing up with the 3.5 versions. One of the large things this hardware will do is support image processing of graphics and images from the camera of a system, and it will allow better pre and post-processing of image operations. Amd has already included the firmware for the hardware, so the plumbing was the last bit that will enable this feature to be utilized or enable the hardware in the CPU to be utilized.
25:00
Finally, this code puts fixes in for some issues with JPEG 5.x firmware loading cleanups. The modules I mentioned before also had some bug fixes squashed in there, general code cleanups and just other bugs squashed in general. It's estimated this should be the last code pull before the 6.11 merge window, which again, you know, a couple of weeks most likely happening. Take a look at the article linked in the show notes for the full details and it includes a link to the official pull request which has all the details of every fix and version number.
25:37
And you know, for people that really like to get under the hood, it's all in there and I'll say I just love the new hardware coming out. I enjoy seeing how much things improve and I also just love to see that we seem to be in a time when we get meaningful improvements. I think some of the maybe little older listeners remember where there was a time when Intel was chugging along with just very minor improvements every generation, and it's good to see some competition. Now we just need Intel to make some leaps so they can really put some fear into AMD and then they can make them compete even harder. So happy computing.
26:15 - Jonathan Bennett (Host)
Yeah, uh, is there? Is there some reading of the tea leaves about when the next AMD stuff is coming out? That they are, uh, that they're prepping for here. When the next AMD stuff is coming out, that they are, uh, that they're prepping for here.
26:26 - Jeff Massie (Co-host)
Uh, not really. I mean, current rumors are saying they it was going to be July 15th, now they're saying the end of July, which would at least get somebody on an RC uh Colonel if they wanted to use some of the image processing that kind of stuff. The current uh Colonel will work with the hardware. It's just you might not be able to take full advantage of some of the special hardware subsystems in there right away.
26:54 - Jonathan Bennett (Host)
Yeah, so there's something here I just noticed that's really interesting and that is AMD's poll touches on the single root IOV SRIOV, and that's pretty important if you want to emulate a second operating system and share a single physical card with it, and that's something that, historically, I don't think AMD has put into their consumer graphics cards. So I wonder if that is specifically their professional workstation server cards cards. So I wonder if that is that is specifically they're they're like, um, you know, professional workstation server cards or if they're thinking about turning it on for for consumer cards now I might be wrong.
27:34 - Jeff Massie (Co-host)
I thought they already had it turned on for consumer cards. I know nvidia will not allow it, even though the hardware is capable. Nvidia forces you to you to get the uh enterprise hardware. But I thought certain cards amd would allow that I don't think on their consumer level cards now.
27:57 - Jonathan Bennett (Host)
The last time I looked into it was a couple years ago, so that may have changed by now. Um, but I don't. I don't know that that's a thing at this point, but maybe it is. Maybe they've enabled it.
28:10 - Jeff Massie (Co-host)
I bet somebody in our audience will be able to tell us for sure, because it's been a couple years. I know I looked into trying to run a virtual machine and set it up so my kids or someone could play off my main machine at the same time. I did, and I have NVIDIA and that's where I found out. Yeah, you can't, there's some gray market ways you can do it or kind of gray area.
28:41
Yeah, like flashing the wrong firmware onto your cards, but don't do that yeah yeah it was pretty, it was kind of sketchy and I'm like I'm not gonna take that chance of bricking a piece of hardware.
28:53 - Rob Campbell (Co-host)
So I just do it. It's only nvidia. You can get a real card. If, uh, if you break it I get more years so looks like after he breaks it, he will.
29:04 - Jonathan Bennett (Host)
Yeah, it looks like as of march 2023, they were amd was pushing in the direction of allowing and enabling sriov support on all of their cards. So, um, yeah, I don't. I don't know if it's actually there yet. Um, I've not seen the big announce that it's possible, but maybe that's, maybe this is more work continuing in that direction. Anyway. So, rob, you want to talk it? I think rob. Uh, rob swapped in his swappy story and uh, I don't. I don't know exactly what that's about. I know what swappiness is and swap, but I don't know what swappy is. It seems like there should be two p's in that yeah, there should be two Ps in that.
29:47 - Rob Campbell (Co-host)
Yeah, there should be two Ps in that. But I'll fix that in post. Fix it in post. So anybody who has been looking at the show notes yes, I changed my story to line up a little better with what Jeff was talking about because my Swapy story is also an upcoming feature in the Linux 6.11 kernel. So in recent times there's been a great debate on swap, swap partitions, swap files.
30:15
When swap is used on a slow drive. You know memory management issues where you know thrashing drives can slow systems to a crawl SSDs if it's not working right or you're swapping too much, you got too much swappiness, you can hurt the right endurance of the SSD. So you know there is just a lot of a lot of pros and cons to swap. If you don't have it, if you don't have enough swap, you can run out of memory. So really having that tweaked properly can be highly beneficial. So in Linux 6.11, they will be offering more fine-tuned control over swap, with a swappiness setting used to determine how aggressively pages are swapped out of physical system memory and into the on-disk swap space. Too much swapping and your drives will be thrashed. So you want to avoid that I've ran into that and done things to avoid that. But now new code from Mark Zuckerberg's meta provides a memoryreclaim that you can configure how aggressive the system will be on reclaiming memory.
31:42
And although I got a little lost in the details on what they were really trying to explain how a lot of it was some of the things I was reading it just I understand swap. I understand swappiness. I didn't quite get how the wording of how they were putting a lot of stuff so I did get a little lost. But there are potential benefits to SSD life, slow-spinning drives, virtual machines and out-of-memory managers such as the SystemD OOMD. So there's been debates. Some people say you don't need swap anymore if you have enough memory. Others say you always need swap anymore if you have enough memory. Others say you always need swap. There's a debate there. I've been in the middle of them. I don't know how accurate or if my side is the official right section way to set up swap and all that, but you know this may resolve some of those problems hopefully that that I have had experience with in a swap yeah, I am.
32:55 - Jonathan Bennett (Host)
I am on team.
32:56 - Rob Campbell (Co-host)
No swap, honestly get enough ram get enough ram, don't turn on swap but I've been told, I've been told by some people, no, you always need swap it's. That's the wrong way to do it. Blah, blah, blah.
33:07 - David Ruggles (Co-host)
It's really, it's really not yeah, I have a lot of ram and even with swap it won't use it, and right, I mean well, that's all configured and, seeing it, seen it not used yeah it'll, it'll use it for some, uh, some processes that, like, haven't been touched for a long time.
33:25 - Jonathan Bennett (Host)
I've seen my systems push a little bit over to swap even though there wasn't particularly memory pressure.
33:31 - Rob Campbell (Co-host)
Yeah, and other systems, like even on Windows and stuff like that sometimes databases or in Windows it's shadow copies, things like that will use up what you have available. You know, if you have 10 gigs available, it's going to use right up to the end. If you have 64 gigs available, it's going to use all of it and then once you start actually doing things, it you know if it yeah, it could still swap if jeff, I believe that well, I was going to say I believe.
34:05 - Jeff Massie (Co-host)
Now, if you go back to episodes 89 and 90, I covered a bunch of the swap file stuff and I'm kind of with Jonathan. There's a lot more people now saying you don't really need a swap file anymore unless you've got extenuating circumstances or you have a very small amount of RAM that you're dealing with and the preferred method now, from what I gathered, was putting it on a swap file versus a separate partition. It lets it run a little better. So if you're really splitting hairs on there, but I'm not running a swap file right now on mine, but I have quite a bit of memory in my machine no swap partition either right in my machine, no swap partition either right, no, swap partition no, but I've got 64 gigs of ram in it, so I'm like, okay, I'm fine, gotta gotta try really hard for that one to get to fill all that up
34:56
yeah, now, swappiness has been around for a while. But I think, looking at the story, this is because you've been able to tune how much it will use the swap file. You know how, how. How much pressure will it put on your ram before it finally says, okay, I've got to trade out to the swap file? But I think the addition in this one is that it's reclaiming that swap file as well. So it will use it. Previously, but the reclaiming was just the normal reclaim. Now you can really say, okay, don't use this unless you really have to. And as soon as you can recover that now you can pull that out much more aggressively.
35:38 - Jonathan Bennett (Host)
So Ilag has a really interesting question here. She says what is the RAM count? What's the limit of RAM for swappiness to be okay? I use 16 or 32. I don't like my second stick. How much RAM do you think you need to have to be able to not have a swap partition or swap file?
35:55 - Rob Campbell (Co-host)
I think it depends on what you do with the computer.
35:57 - Jeff Massie (Co-host)
Yeah, it's really going to be what you know. If you're just reading email, you're 16 is going to be fine Surfing the web, Large image manipulation, things like that.
36:07 - Rob Campbell (Co-host)
That's where you start reading email Unless you have hundreds of browser tabs open. But even that, I think most browsers have features to put to sleep.
36:17 - Jonathan Bennett (Host)
Yeah, they put the background tabs to sleep now.
36:23 - Jeff Massie (Co-host)
Right. So at worst just use your 16 and just see what happens. At worst it's going to crash. Now you could have a swap file crank up the swappiness really high so that it's very aggressively not using your swap file, and you can monitor that over a longer period of time to find out if you've ever touched a swap file. I've run smaller amounts of RAM and have not run into issues.
36:52 - Jonathan Bennett (Host)
Yeah, One of the things with swap that can get you is if you're running a conventional hard drive it's so slow. And then on solid state drives, unless you've got one that's made for it, you run into problems. Now, from what I understand, this is not as much of a problem now as it used to be, but you run into the wear problem on SSDs. If you have a machine that's really swappy and you can eventually burn out an SSD.
37:19 - Rob Campbell (Co-host)
Yeah, the problem I've seen is I've had systems where they're just slow and I look at it and it's the IO, and I come to find out that it's a swap that's just making it drag and that's spinning drives.
37:38 - Jeff Massie (Co-host)
Well, even with SSD it's exponentially slower than RAM, so it's still going to slow you down. But yeah, the burnout of ssd is not not near what it used to be.
37:50 - Jonathan Bennett (Host)
I mean, you can, you can run them and thrash them pretty hard these days and they automatically kind of self-check and some of those because they've got they've got automatic wear leveling built into them and I think, I think ssds these days are shipped with unused blocks to be able to swap out when, yep, when they start hitting the way levels.
38:07
Yeah, all right over provision, just for that very reason so I did want to touch back on the uh, the sriov issue. I did a little bit more looking and it looks like amd for sure. That is only on their pro cards. None of their consumer cards at this point have it. People have been asking for it for a long time and a quick look shows that on NVIDIA it is probably the same thing. That is something that's only on their pro cards. Intel, interestingly, has it turned on for just about all their cards, it looks like. So there you go. Let's talk about Nexus mods. Ah david, what is going on with nexus mods and the steam deck?
38:52 - David Ruggles (Co-host)
well, it's specific to the steam deck, at least the way it's written, but, as we all know, steam deck is just fancy linux and we all use linux, so we should let our voices be heard as well. The Nexus mods have a brand new app that they've put out. It looks quite promising and exciting for the future of modding, especially for Linux and Steam Deck, and now they want your feedback. So their plan with this is to replace previous apps like Vortex, and right now it's only ready for Stardew Valley, since that's a very popular game. But they announced the alpha release of the next generation mod manager and their new product manager got in touch to mention they would be really keen to get feedback from Linux users specifically really keen to get feedback from Linux users specifically.
39:46
So this is your chance to ensure Linux and Steam Deck finally become a first-class citizen for game modding. We've talked a lot about game support and all of the work that's being done for that, but not as much about the mod community around that. So they do plan to release updates about every three weeks, and the next game they plan to add support for will be Cyberpunk 2077, which is one that's on my list to play, but I haven't touched it yet. So if it's something that applies to you and you game on Linux, reach out to them. They want your feedback.
40:22 - Jonathan Bennett (Host)
Yeah, that's interesting. So I've done a little bit of modding. Nexus mods is one of the big names in the modding scene for for a lot of different games and it can be a pain. I've done I've done some modding on unity games and getting that to work, trying to get, you know, a unity mod install installed in the same you know the same bottles where the game is, or pointing it at the right place. It takes a little bit of fiddling to get a Unity mod install installed in the same bottles where the game is, or pointing it at the right place. It takes a little bit of fiddling to get it lined up. So I can see some definite potential here. On the other side, though, looking at this, one has to ask themselves is this just spyware that also happens to install video game mods? Wow, is this just spyware that also happens to install video?
41:08 - Rob Campbell (Co-host)
game mods Wow I didn't see the word AI in it.
41:10 - Jonathan Bennett (Host)
That's true. That is in their favor.
41:18 - Rob Campbell (Co-host)
That is in their favor and David Jeff could tell you all about Cyberpunk 2077. He's a big fan. Okay, we'll take that offline. No, I'm joking.
41:28 - Jonathan Bennett (Host)
I played it, but I wasn't so much of a fan, I see potential with this. It's neat to see a business coming along and trying to give it some support. Hopefully it'll make things easier. We'll have to figure out how to run it safely. We'll tear it apart and see what all it's sending back. But yeah, it's going to be cool.
41:49 - Jeff Massie (Co-host)
Yeah, I only had one game I was not able to successfully mod on Linux and that was Dark Souls 3, but that's because it replaced a Windows system file and so it kind of messed with some of the library that Wine was looking for and that was one that just it was not happy.
42:11 - Rob Campbell (Co-host)
Are we talking about actually using mods on Linux games, or are we talking about being able to do the modding yourself with this?
42:22 - Jonathan Bennett (Host)
It's installing of mods.
42:24 - David Ruggles (Co-host)
It's a mod manager.
42:26 - Jonathan Bennett (Host)
Other people have developed mods. I'm sure you can test with it, but this is not going to be for developing mods. It's going to be Nexus mods. It's a repository for a lot of people to host their mods. I use it for the Pathfinder games. In fact, there's some fun stuff that happened there. The original Pathfinder tabletop game is turn-based and the first pathfinder video game came out and had no turn-based mode and everybody's like, oh, this kind of stinks, and the people that made it were like who would want to pay it, play it in actual turn-based mode? You couldn't even do that. So somebody wrote a mod that made it turn-based and put it on nexus mods and it was great. And so the second Pathfinder game came out and officially had a turn-based mode and they credited the mod makers. They're like this guy showed us that it could be done and it was great. So we did it. Um, do you get all kinds of fun stuff like that on Nexus mods? Just crazy bonkers mods where people add new stuff to games.
43:20 - Rob Campbell (Co-host)
it's great. I'm not familiar with this, so it sounds like it's like a repo, like a software center, but for mods. So this isn't for the developer, this is for the general gamer. Yes, so they can. Just, you know, I like developing but I don't have much time for it these days, but I do have time for games for some reason. So this would be for me. I could get this and I could go in the stores like, oh yeah, let's get that mod and that sounds fun, and get that mod and boom, I play it and I get the mod. That's cool.
43:53 - David Ruggles (Co-host)
Yeah now to circle back around on that. They are only supporting. What did I say? It was Star Valley.
44:04 - Jonathan Bennett (Host)
Yeah, just a very small list of games at this point.
44:08 - David Ruggles (Co-host)
Yeah, because it's alpha release, so they start with a single game and then the next game they're going to add support for is Cyberpunk, which you already said you didn't like. I don't know that you could use this right now but it's coming.
44:20 - Rob Campbell (Co-host)
I just had a hard time getting into it. I don't know what it was. Also I, I just had a hard time getting into it. I don't know what it was. So also, I'm assuming this is like the Linux version of the games and not like the Proton version, since that would really be a Windows app.
44:35 - Jonathan Bennett (Host)
See, that's an interesting question. I don't know, I don't think there is a Linux cut of a cyberpunk 2077. Is there?
44:45 - Jeff Massie (Co-host)
No, it's, it's the proton version. But see the, what it's doing is it's installing it in the path where you, where it needs to find these, but the paths are different because of being in wine. So it's it's got to kind of refigure where everything goes and some of the mods are rather small, like I remember, like Witcher three. I got to kind of refigure where everything goes and some of the mods are rather small, like I. I remember, like witcher 3, I had one that would just like auto loot for you, you know, so you don't have to click on everything all the time. You know, some of them are just like. One of the famous ones was, uh, one of the doom games. They had a duct tape mod. They're like I'm in space and all this and I can't have my weapon up and a flashlight. Where's my duct tape?
45:25
tape the flashlight for the weapon you know it was so. Sometimes they're just small little things.
45:33 - Jonathan Bennett (Host)
They can be as trivial as adding extra uh art, like um character art, avatars to the game. You know, you, you want to be able to use either custom or, you know, add more, um more, more pictures to be able to use, either custom or add more pictures to be able to use for your character, all the way up to completely game-altering mods, like the turn-based mod for Pathfinder, which was great. All right, let's move along. This is not the gaming show, although we talk about gaming a lot. But let's talk about Fedora. We talked about spyware. Let's talk about other potential spyware. Jeff has the story about Fedora. We talked about spyware. Let's talk about other potential spyware. Jeff has the story about Fedora Fedora adding spyware. I tell you what. What's the world coming to? Oh, it's not quite that bad.
46:17 - Jeff Massie (Co-host)
Not everybody needs to jump to Debian, but we'll go through it. So Fedora is looking to put opt-in user metrics collections in fedora workstation 42. Now that sounds kind of familiar. That's because they tried it before in fedora 40 workstation. I should be clear that workstation is the only version that was affected. No, you know, none of the other versions were included and but it you know that was a major issue for a lot of people.
46:46
Now you could control if the data was going to be collected, but it was opt-out, so that means it was on by default. There was a switch to turn it off but, like I said, the setting was defaulted to on and there was a very large uproar about that at the time and really in general, linux users are not too keen on data collection, especially if it's defaulted to on and if you do anything on Linux this is my editorial here you need to be really clear on what you want to collect and how it will be used, and it needs to be defaulted to off so every user can decide if they want to contribute to analytics. Now the analytics they were collecting didn't have any personal identifiable information, so no one was going to be able to figure out who you were from the data, though I think the data collection that was going on in Windows at that time also had people a little sensitive to information collections, because at the time Windows was really ramping up their stuff. So the computer audience in general is very sensitive to any kind of collection. Now fast forward to Fedora 42, and I do want to be clear this is a proposal and is not implemented yet. So in the links in the show notes they have links to the Fedora wiki and it goes over all the details. They do state that this is the Fedora wiki and it goes over all the details. They do state that this is you know, and this is directly from the wiki, though I'm going to summarize it some for time. So protecting user privacy is of utmost importance for this initiative. To this end, the service will only collect generic, standardized data and will never collect anything that is personally identifying. It will also, of course, be fully open source. On the server side, the data will be stored in a way that prevents user identification. The data collection will be run in a transparent manner and will be governed as part of the Fedora project. Fedora intends to ensure that the metrics reporting is fully under the control of end users. Metrics collection will default to off and will only be enabled through a clear on-off prompt in initial setup. Users will be able to view the data that has been collected locally and will be able to remove the client software from their systems, should they choose to do so. Okay, well, that sounds well and good. What are they planning to do so? Okay, well, that sounds well and good. What are they planning to collect?
49:07
Now, in the wiki they have categories with some examples. So just, for example, hardware details. So that would be CPU, graphics cards, cameras, which peripherals? Do you have things of that nature? It's the Steam survey. Yeah, the Steam survey System settings the display language, which input methods are used, which accessibility features are enabled.
49:30
They have desktop usage patterns, which apps are used, how many open workspaces there are, how often each system settings panel is opened. They have performance reports, so disk and memory usage, and then evidence of problems, you know, counts of system crashes, out-of-memory events, app crashes, things like that. Now they also list their steps on how they're going to ensure being anonymous. I won't go through it all but, for example, like the app usage, they will only collect known packaged apps and anything custom will not be collected. So these are would be apps that would be in the repositories or, you know, on FlatHub, things of that nature that are well known out there apps. You know, a bit of editorial for me.
50:17
I think this will work out okay for them. They're being very transparent on what they're going to collect it. You know it will be off by default and they have a level of transparency that was not used before by Fedora. So this is a lot more transparent than the initial offering that they had and they do talk about. They listen to the user feedback and that's why they made these transparency changes. There's a lot of details in the wiki. They even cover how you're going to request copies of the database, they're going to show initial examples of the data before it fully goes live, and they even cover the process they need to go through if they're going to make any changes, which boils down to they're going to let everyone know and say why they need to make the change and get a chance for feedback. So part of the community ability to talk on changes. So for me, I'm okay with this because it's how open they're being and even looking at comments from the community on many different forums I would say about 90% are in favor of this change and are okay with it. Some say they won't ever turn it on, but there are actually quite a few that don't like telemetry. But because of how they're going to do it, how they're even thinking it will turn out to help Fedora, they're going to turn it on. So it's the openness and complete sharing of data seems to be really endearing themselves to people who are maybe on the fence or would default to no.
51:53
Now I should state this again this is for workstation only. The other versions are not getting this. So if you're running regular Fedora with KDE or whatever, this doesn't apply. Workstation only Now. You know, maybe in the future, but nobody's even talking about that right now, only talking about workstation. So you know, take a look at the show notes for links to the articles. Read the wiki, because it's fairly lengthy. I mean not terribly so, but I mean they really go into a lot of very specific details, answer a lot of very specific questions about how all this will be collected and how, how they will handle different uh scenarios.
52:35 - Jonathan Bennett (Host)
So you know, let us know what you think on the discord server so the way, the way they describe this, it's just going to be another. It's going to be another package, it's going to be another RPM package, so there's nothing that would keep you from installing it on a server install. You might not want to, but there's no technical reason why you couldn't, which could be interesting.
52:57 - Jeff Massie (Co-host)
It will default to installed, but it will not actually be operational.
53:04 - Jonathan Bennett (Host)
So I'm just saying there's nothing that would stop a user from installing the package on a server install and then going in and turning it on. Now you may have to install all of GNOME to be able to get into it and turn it on right, like there may be a huge dependency, but you can. The only this is what I'm saying the only difference between Fedora Server and Fedora Workstation is the list of packages that are installed um, yeah, so so you could.
53:28 - Jeff Massie (Co-host)
It's just like you said you're going to probably have some dependencies, you're going to have to load into yes to make it seem like workstation.
53:34 - Jonathan Bennett (Host)
Yep, yep, um. I I appreciate the way that they're going about this. Now I saw I saw at least one one of the devs that said he was against it and his comment was if you leave this, what they're doing is they're going to have like a pop up during install. Here's what we want to do. Nothing is selected, click yes or no. And the comment that he made was users are just going to click yes without reading it and install it, and you know it and like I hope linux users are a little bit more sophisticated. I'm like he's right in a way, but at the same time, you don't get to be a good linux sysadmin by just clicking yes on everything.
54:18 - Jeff Massie (Co-host)
That's just not how that works um, well, especially if it's workstation, the audience is going to be a more advanced. You would not think, grandma you know it's not.
54:28 - Jonathan Bennett (Host)
It's not a windows install. Um, yeah, yeah, I, I don't. I don't have a problem with it, so long as it is done in the open, it's not a hidden thing, and it's not installed by d or it's not enabled by default. Let's say, say, and on top of that, easily uninstalled, I don't have a problem with it.
54:47 - Rob Campbell (Co-host)
I kind of suspect iLegs. Well, ilegs said you should click no without reading and you know really, I think, what a lot of Linux users I want to say the majority probably- are like that it's like yeah, nope, yeah, I think there's's gonna be a lot more no's than yeses.
55:06 - Jonathan Bennett (Host)
I would I would agree with that. I mean, that's what I do whenever I install most anything. Can we? Can we collect data? No, thank you, please don't can we collect data.
55:15 - Rob Campbell (Co-host)
Do you want to install this? Do you want to install that? Do you want us to do this for you? Nope, nope, nope. I just want to install it.
55:23 - Jonathan Bennett (Host)
Don't upsell me on anything yeah, goodness, I just about use the fedora minimal installs these days like for anything other than a full-on desktop workstation. It's like nothing extra in there. Let me. Let me choose the packages, things that I want turned on. Um, yeah, she, she says every website is. Uh, please let us put cookies on your computer. Decline only the necessary ones that is understandable.
55:54 - Jeff Massie (Co-host)
You know, I actually uh do turn on some of that uh um data collection stuff, depending on which uh project it is and if I'm willing to help them I don't know how I?
56:08 - Jonathan Bennett (Host)
yeah it, so it depends right? So I'm trying to think of an example of where I turn it on. There have been some where I've turned it on, but I have with kde yeah, I may have with kde I don't remember and kate.
56:20 - Jeff Massie (Co-host)
Now kde actually has a slider. So you can, you know from most basic information to give a little more specific. I mean it's still pretty generic, but you know, I think when you said it's like than we thought, hey, we better take that into account when we spin the next distro to make adjustments.
56:51 - Rob Campbell (Co-host)
Yeah, and for me I said most people are going to click. No, I'm not like that. I mean, when I'm using Linux, these are projects I care about. I care about the software I'm using. Therefore, I want to help them. For the most part, I'm not giving them money sometimes, but mostly I'm not giving them money. So I want to help them, show my support, show my numbers, like, hey, I'm here, please port your software because I'm here, kind of like I say every time I bring up the Steam server, I'm like vote. Often I definitely don't skip that. I want my voice to be heard. I want people to know hey, the numbers are here on Linux, we're using it. Count me, track me. I mean.
57:39 - Jeff Massie (Co-host)
And there's some of the, and there's more of us, you know, maybe using ultra-wide monitors or something. Hey, take that into account.
57:47 - Rob Campbell (Co-host)
Just you.
57:49 - Jeff Massie (Co-host)
Just me, but I voted Right, so that means a lot.
57:53 - Jonathan Bennett (Host)
So that's something to think about. So developers are more likely to put time into things that people are actively using, and so you run your application, let's just say KDE. Okay, you turn on the analytics for KDE and it sends back to the KDE folks that this person uses Wayland rather than X11. This person uses the console. This person also uses Uquake. This person really doesn't use k mail, though, and so they go oh well, we need to make sure that those things are up to date, and maybe we don't have to put as much resources into k mail. Eventually, enough people do that, and they're going to start thinking about is it time to retire k mail now? I'm just I'm making this example up out of whole cloth. You did not hear it here. Kde is not looking to retire kmail. You can keep your kmail. That's not the point, um, but it can be useful to developers to know what parts of your, your ecosystem, people are actually using all right, unless, uh, jonathan gets it right.
58:59 - Jeff Massie (Co-host)
Next week we find out k mail's going away and then it'll be like, okay, every show, jonathan, give us the weekly lotto. They're listening.
59:07 - Rob Campbell (Co-host)
Well, it's time to retire K-Mail.
59:10 - Jonathan Bennett (Host)
Yeah, no, no, no, no, no. All right, so let's talk about Proton, Proton again, proton again Same Proton, different news, different news.
59:25 - Rob Campbell (Co-host)
So for privacy-focused Linux users, most are familiar with things like ProtonMail, ProtonVPN. It's not exactly a Linux, it's not exactly Linux but it's very Linux-adjacent having these things, and Linux users are kind of familiar with it. But if you're a Google Docs docs user, um, proton still wasn't about to fill that need for you and and get you out of the google verse. This week proton announced docs in uh proton drive, a privacy first document editor which allows anyone with a free or paid copy of Proton account to create, edit, share and collaborate documents on the go from any web browser. Proton Doc supports rich text editing, tablet support, paragraph styles, image embeds and file compatibility with common formats like docx and rtf. But disappointingly for us open source Linux and LibreOffice users, it doesn't support ODF, the open document text format that's used by default in LibreOffice. But like Google Docs and Word Online, it supports real-time collaboration with others you share documents with, including inline indicators of whose caret is where and the ability to make comments and replies all end-to-end, encrypted and secure. Obviously it's Proton, so it secures their focus here, but it's not a comprehensive word processor and lacks many formatting advanced features found in Google Docs, zoho, onlyoffice, online editors, etc. So for writing reports and college papers, you may want to stick with something more like LibreOffice for now, but it's new and hopefully will improve and still seems very promising, you know, for maybe something like our show notes. It could be useful and we don't need all that fancy editing. But that's not the only thing.
01:01:40
Proton is in the news for this week. This week they have announced that they are transitioning into a non-profit. Proton has already been a supportive member of the tech community, giving grants worth more than $2.7 million to advance online freedom and democracy around the world. But adopting a Swiss nonprofit structure provides additional security, which a corporation cannot achieve. Because Proton has no venture capital investors. They could take this additional step to secure that future.
01:02:16
Swiss foundations do not have shareholders, so Proton will no longer be dependent upon the goodwill of any particular person or groups of people, allowing them to focus on their mission to remake the Internet in a way that is private by default and serves the interest of all of society, not just the interest of a few Silicon Valley tech giants. This is in their words, not just the interest of a few Silicon Valley tech giants. This is in their words Proton wants to create an internet that is able and willing to defend freedom no matter the cost, and they believe being a nonprofit is more aligned with their mission. So you got Proton Docs and you have a non-profit uh support on those docs. So you know, with the google docs it's not no non-profit there and not security first so there is one.
01:03:16 - Jonathan Bennett (Host)
There's one thing in particular that I note about what proton is doing with their non-profit. That is very interesting. It is going to be a swiss non-profit. I can tell you why. It's so that they they have a place to retreat to if they get one of those fancy national security letters or the equivalent in other countries. Oh sorry, we're a swiss corporation, we're not subject to that. We don't have any servers in your country anymore well, isn't proton a swiss company anyway?
01:03:45
they were started by some folks from cern and they, they may, you know, they may have always been based in switzerland. I don't. I don't know that for sure. I just I find it very interesting that that's where they're at. Yeah, of all the places around the world, them trying to be um, um, uh. Help me out here, guys. What's the word I'm trying to think of? You're supposed to be my backup here. Um, neutral, country, neutral, thank you.
01:04:15 - Rob Campbell (Co-host)
Ah, my tip of the syndrome gets really bad some days I mean yes of all the places for them to be, for them to be neutral, switzerland is the one that makes the most sense I mean I think a lot of that makes sense too, with all their privacy focus, their privacy focus, vpn and proton mail and now their proton docs and their drive and all that.
01:04:34 - Jonathan Bennett (Host)
I think it makes a lot of sense being there and having that security focus, just kind of it's that neutral security blanket, yeah, although, let's be honest, if you really really needed privacy on your documents, you need to run an xCloud on your own server.
01:04:58 - Jeff Massie (Co-host)
That's really where it's at.
01:05:02 - Jonathan Bennett (Host)
And write it in cursive on paper. Nobody would want to read it then. All right, let's chat about more about the kernel. Uh, get random, this is. People are talking about it being another epic torvalds rant. It actually it's. It is, but it's not. Um, it's actually a neat of code that lots of things can get to, and I think it's used for virtual machines as well. They were looking at adding get random into that, basically making get random a system call that you could talk directly to the kernel to. Making get random a system call that you can talk directly to the kernel to.
01:05:45
And a friend of ours, jason Donnefeld, is the one pushing for this, and Torvalds kicked it off in the words of Michael Larrabelle. Torvalds kicked it off by writing nobody has explained to me what has changed since your last VDSO get random, and I'm not planning on pulling it unless that fundamental flaw is fixed. Why is this so critical that it needs a VDSO? Why isn't user space just doing it itself? What is so magical about this at all? Et cetera, et cetera. He says it seems entirely pointless to me. I was very skeptical last time and nothing has changed. He says that we don't add stuff just because we can. We need to have a good reason for it. And I still don't see the reason Torvalds goes on to say in other words, I want to see actual users piping up and saying this is a problem. This is my real load that spends 10% of the time on get random that fixes it. He says I'm not at all interested in micro benchmarks or theoretical. If people need high performance randomness, I need a real, live user that says I can't just use RDRAND. And then he talks about. He says I'm so negative about this all because this is interesting stuff. The random number subsystem has such absolutely horrendous history of two main conflicting issues People wanting reasonably usable random numbers on one side, and then people that discuss what the word entropy means on the other, and honestly I don't want the kernel stuck even more in the middle of that brass. So, oh, one more, one more. Jason, this smells talking to Jason Donofill, this smells it's BS. Let's make a deal Do a five-liner patch that adds the generation number to the VDSO data documented as the kernel thinks you need to reseed, and if it turns out in the future that there's any major reason why it doesn't work, I'll take the thousand line thing. Okay, deal.
01:07:37
I went and read and I would encourage actually everyone to go and read the exchange on the Linux kernel mailing list. It's actually pretty interesting because Torvalds is not holding back here. He thinks this is a terrible idea, but Donofield responds back and says no, no, no, this is not the magical we want to fight over randomness thing. There is actually a real problem here, and a couple of other users do chime in, and what really interested me the most is the thing that they pointed to that could be a real world problem, and it's not performance. What they pointed to is if you do a clone of a virtual machine so we have a virtual machine running and you say, all right, pause it, get a copy of its disk and a copy of its memory and start both of them side by side, which is something you might do for debugging purposes it is you run into problems where, like if you generate a UUID, a universally unique identifier, on both machines at the same time, which you might do to generate a log line it's going to generate the same UUID because its randomness pool is the same. So this is apparently a real problem that shows up when doing debugging on virtual machines.
01:08:52
And there was another couple of issues that came up and Torvalds eventually comes back and says I guess I'm going to have to walk through the patch series once again. I'm still not thrilled about it, but I'll give it another go. And the last message that I read on that thread just before the show started, he goes, I don't. He points out something. I really don't like this, but the rest of it's okay. Guess, so it's just it's. It's such an it's so interesting to see the back and forth on the kernel mailing list. Between you know, torvalds thinks it's a terrible idea. Somebody explains where it's not not a terrible idea for all thinks about it. Okay, maybe it's not a terrible idea, it's just a bad idea. Okay, maybe it's not such a bad idea. I guess we can pull it like. It's just fascinating to me. I I really enjoyed watching the back and forth, how it goes yeah, that, that's always interesting um, to dive into.
01:09:40 - David Ruggles (Co-host)
You know the thought process. Okay, so to take a step back, that's one of the things I love about open source in general and, um, just that, open dialogue and really brilliant people that are 10 levels above me. Stepping through stuff helps me learn and absorb more knowledge.
01:10:01 - Jonathan Bennett (Host)
Yeah, absolutely. I very much enjoy kind of the bizarre method, not B-I-Z but B-A-Z-A-R, you know, as opposed to the cathedral, the bizarre method of software development, especially when it's open source software, because you do, you get to see how it's made. If you go and you look for it, you can get insight into why decisions were made. You can even give feedback on it, and a lot of times that's useful.
01:10:28 - Jeff Massie (Co-host)
I think one of the interesting parts too is seeing Linus kind of backing off his stance when people start. Really, you know, he's not totally inflexible. If you can make a good argument you can change his mind and to me that shows you know being a good leader, you know being able to say well, here's how I feel. Oh, here's a bunch of data, oh, maybe I was wrong.
01:10:55 - Jonathan Bennett (Host)
Yeah, so I was thinking about that just now. Obviously, torvalds has an a masterful grasp of the colonel, although he will. He has admitted that there are parts of it that he doesn't understand everything that's going on, because it's not what he works with all the time. Um, but the neatest thing about torvalds, and one of the best things about a good manager, good leadership, is it is that idea to change your mind and rapidly change your mind too, and I would say that's probably more important even than a technical mastery is being able to kind of weed out what's accurate and not about what people are telling you and then admit that you were wrong about it, and that's that's boy. That's just about the thing that the colonel's got going for it the most is that they are, they could, they could pivot, like all the people behind it. They can pivot if they need to. It's pretty neat to see. Let's chat about kde. Then kde plasma 612 is out. Uh, we are looking forward to 6.2. David, you want to take it away and tell us about that?
01:11:58 - David Ruggles (Co-host)
Well, kde may not be pivoting, but they are plodding forward. There you go 6.1.2 is out, and it is out only a week after 6.1 was released. It sorry 6.1.1 was released. They're continuing to address those little nitpicky issues fixing bugs, improving KDE and its integration with Wayland and all the other window management stuff. So 6.1.2, it makes titles more legible. They added some bell sounds to some of the sound themes, fixes, hover and highlight effects in the sidebar, in specific section of accessibility pages. System settings correctly hide some subcategories. Just a whole bunch of different fixes. It did address a regression which we've already discussed the word regressions that means something that was fixed and then got unfixed, and now it's fixed again which broke the layout of Plasma Discover's share dialogue. Plasma discovers sheer dialogue, a K-Win crash when run on a system with graphics hardware older than 15 years and a memory leak in Plasma. So there's a full change log in the show notes or in the article linked in the show notes if you really want to dig into it.
01:13:31 - Jonathan Bennett (Host)
But it's always good to see progress being made yes, I went ahead and added the the link to the weekly pointed stick. That is uh, that is nate, uh, nick graham, uh, the nate's blog about the things going on and he's talking a lot about what's coming in 6.2 and they've got some fun ui improvements and fixes and all kinds of stuff there as well.
01:13:55 - Rob Campbell (Co-host)
So yeah, you really caught my attention with the adding of the bells. That's going to be awesome.
01:14:03 - Jonathan Bennett (Host)
The important stuff.
01:14:05 - Jeff Massie (Co-host)
I will say, you know, I'm really impressed with KDE 6. I mean they're aggressively, you know, getting it up to speed. I mean it's a lot of changes quickly stomping bugs, adding features.
01:14:20 - Jonathan Bennett (Host)
One of the things that just landed in the last time I updated so it was either 6.1.1 or 2, I think was they went in and added the option to pass mouse clicks through to old legacy X Windows clients. And you might think, well, that sounds weird, what's going on with that? Well, that is for those of us that have fancy mice, that have extra buttons and we want to be able to do things like push to talk with the fancy buttons. And some of those applications are not up to stuff for Wayland. They're still using X-Wayland, and so now I can run mumble and I can use my fancy button on my fancy mouths as my push to talk, and that makes me happy. It's very niche, but it makes me happy.
01:15:12 - Jeff Massie (Co-host)
You know it'll be interesting to see because for a long time with kde they said oh, you know, you always waited until the 0.4 release before you you jump to it because that was where it was. It was pretty stable and feature parody from the previous version. Curious if it'll be that or it's kind of already there. It seemed like what I've run so far.
01:15:32 - Jonathan Bennett (Host)
It's already pretty good. Yeah, kde 4 was the scary one. Kde 4 was the real troublemaker. That was the one where you needed to wait several versions and I think they kind of learned the lesson from that. Like, let's make sure it's a little bit more baked before we call it the .0 release.
01:15:50 - Jeff Massie (Co-host)
Well, they partially didn't advertise that very well either, because when kde 4 0 and 4 1 came out, it was never really meant for everybody to jump to it. It was still in kind of that beta-ish stage, yep, but they didn't really say that very clearly. You kind of had to look, look in the corner and under the covers to find it. Yeah, yeah, they, yeah, they got a black eye on that one A little bit yeah, All right.
01:16:16 - Jonathan Bennett (Host)
Well, let's talk about ooh, let's talk about Bolt and Meta Meta. What is Meta doing to the kernel? Why is Facebook messing with the kernel?
01:16:26 - Jeff Massie (Co-host)
Well, you know, meta has a mixed bag of feelings for a lot of people, but when you look at the open source, they actually do quite a bit. A mixed bag of feelings for a lot of people, but when you look at the open source, they actually do quite a bit. Now you can argue that they work on things that help themselves, but I would argue that's what most companies do, that's what all of us do, yeah, with open source, though we can all share the improvements, so it's all good.
01:16:45
In this latest bit of code they're working on, they're looking to optimize the kernel with bolt, which, as the title in the article linked in the show notes says, is getting a 5% improvement. Bolt is a way to speed up Linux ELF binaries and it stands for Binary Optimization and Layout Tool. Bolt works with the LLVM compiler and something that the kernel needs to be built with. This isn't just with the LLVM compiler, I mean, it isn't just an option to toggle with the binary. It's been proven in the past to have a speed increase in user space applications, but now they're focusing on the kernel. The LLVM documentation. Talking about Bolt, which is linked in the article in the show notes, has this to say about how it works, and I'm quoting here Bolt optimizes code layout based on a low-level execution profile collected with the Linux perf tool. The best quality profile should include branch history, such as Intel's last branch records, lbr, such as Intel's last branch records, lbr. Bolt runs on a linked binary and reorders the code while combining frequently executed blocks of instructions in a manner best suited for the hardware. Other than branch instructions, most of the code is left unchanged. Additionally, bolt updates all metadata associated with the modified code, including Dwarf, debug information and Linux or unwind information. So this is a multi-step process, and how this works Now this is going to be greatly simplified, but the process looks like this.
01:18:23
So like first, you compile the kernel like you normally would, except now it will create extra files and one of them is an uncompressed Linux image. Normally, you know, when you build a kernel, Linux gets a compressed image and they're decompressed into memory at runtime. But we'll have an uncompressed version this time. You use the perf tool, performance tool, and start your workload and they give an example such as a database benchmark, and you collect the data while the system's under load. You convert the data collected to a format that Bolt can use and you can merge data from several workloads in at this point if you wanted to and you make a couple changes to reserve space for Bolt and run a patch that runs on the kernel binary for Bolt and run a patch that runs on the kernel binary. So you know, as you can see, bolt, while giving a nice performance improvement, is not set up for general purpose computing. So this is not just oh, I'm going to turn this feature on and it's going to work great. It's pointed more towards servers or enterprise machines which run very specific types of workloads, such as a system that runs a database with a high load and the system is only going to run that database. Or, you know, there could be a few other applications as well, but you're not going to run performance across everything you're doing Now.
01:19:46
By making some small changes, bolt can help reduce the amount of instruction cache misses and wrong branch predictions. So that's basically making sure that it's more accurate in its predictions Not that they still won't happen, but it will be with a lower frequency. Now it's also going to depend on how much time the program is in the kernel. If there's just a little bit of time, you're not going to see that big of an improvement, but an application that hammers a kernel all the time will see a large benefit. Now take a look at the article in the show notes and for those of you who handle servers that are bottlenecked with the kernel, this will help give a little more boost to your workload. Notice I said kernel and not hardware. It is possible that you're hardware limited and this won't help very much. This improves kernel bottlenecks, so take a look at if you have. If you're running kind of a enterprise or special application, that's a fairly heavy load on your server and maybe this will speed things up.
01:20:45 - Jonathan Bennett (Host)
Happy compiling yeah, yeah, it's interesting stuff, it's. It's pretty neat that the way they're doing this is you feed your workload into it, you run your workload for a while and then it uses that data collected doing your particular workload, to do the optimization. Like that's pretty specialized optimization, but if you're running a server farm then yeah, it makes sense.
01:21:08 - Jeff Massie (Co-host)
Yeah, if you can get 5% more out of 1,000 servers or something that's.
01:21:13 - David Ruggles (Co-host)
That's suddenly a big difference.
01:21:15 - Jeff Massie (Co-host)
Yeah, you're saving money on fewer servers and power and whatnot.
01:21:21 - Rob Campbell (Co-host)
What if I have a VPS and I'm trying to save money and have as few resources as possible? Could this help me get that a little more, or is it just going to be so incremental that On a single VPS probably not going to make a difference.
01:21:39 - Jonathan Bennett (Host)
If you've got 100 servers running virtually, then it might.
01:21:44 - Rob Campbell (Co-host)
No, well, I got one. It's not going to help.
01:21:48 - Jeff Massie (Co-host)
Are you kernel limited? So you need to use your perf tool to find out where are your hangups, if you're. If you're, yeah, if you're. If you're hanging up in a user land now, bolt can help there, but you can have to profile it, and if you're hardware limited, it's not going to help you. Yep, it's going to eliminate your you know cash misses, your wrong branch predictions, things like that, or reduce them, I should say, not eliminate yeah, interesting stuff.
01:22:20 - Jonathan Bennett (Host)
Um, so I think that's the news. I think that's it. Let's get into some command line tips. Not all command line tips, um, I see at least one that's a website, so that's probably not a command line tip. David didn't get the memo I did again this week. Ah, good job.
01:22:39 - Rob Campbell (Co-host)
We'll start with rob, though, and what is uh, inksy well, um, I don't know what, how you pronounce it, but but, uh, I just call it I-N-X-I myself because, uh, I don't know Inksie, whatever. So, anyway, uh, for those. So this is a fairly new command to me. I have not heard of this before, I don't know a few weeks ago, but it seems like recently. I keep hearing, uh, or seeing people mention inksy, uh, specifically usually on like support forums and things like that. People are like, hey, I'm having this problem like run inxi or in yeah, inxi, dash this and show us what it says. And I don't know, maybe it's to replace a NeoFetch, I don't know, but here's what it is, here's what it is. So INXI with a dash B is going to give you basically system information. So for those watching you can see what's on my screen. No, you can't, because it didn't update, did it? My screen is not updating. Let me do a quick little.
01:24:03 - Jonathan Bennett (Host)
Refresh your OBS.
01:24:05 - Rob Campbell (Co-host)
Yeah, let me quick refresh my OBS here, because it's being dumb.
01:24:10 - Jeff Massie (Co-host)
It's on the other monitor.
01:24:11 - Jonathan Bennett (Host)
That could be.
01:24:16 - Rob Campbell (Co-host)
Oh no, I don't Let me see here, hold on.
01:24:19 - Jonathan Bennett (Host)
All right, here's what we're going to do. We are going to go to David and we're going to get his command line tip, and then we're going to come back to Rob, who will see if he can work his issues out. So, david, what about Roll Dice? So, david, what?
01:24:30 - David Ruggles (Co-host)
about Roll Dice. Roll Dice is a cool little command utility for all you tabletop gamers or anybody else that needs to roll dice. I confirmed that it is in the repos on Ubuntu. I think it's installed. You simply type roll dice, all one word, space the number of dice, lowercase D and then how many sides the dice have. So if you're doing a traditional six sided die for a game of Yahtzee, you can do that. Or if you're deep into Pathfinder and you need a 20 sided die which is a guess, because I don't actually I've never played Pathfinder.
01:25:15 - Jonathan Bennett (Host)
It's 20-based yes.
01:25:17 - David Ruggles (Co-host)
Okay, so then you can do like 2D20, and it will do 2D20 dice. There are a lot of different commands. You can do all sorts of modifiers and things automatically, and one of the interesting options is a dash s on the command line which will separate. If you're doing multiple dice, it'll separate them out, or without that, it just gives you a sum of the total. So if all you need is just a total score, then you don't need the dash S. But if you need to know the values of each individual dice or die in the dice roll, you can do a dash S. There are other options to give you full verbosity and show all the steps it takes. It can be run interactively so that you don't have to exit back out of it if you're going to be doing it repeatedly. So it might be a little bit of overkill, but what are geeks for if not for technical overkill?
01:26:21 - Jonathan Bennett (Host)
So, anyway, I wish I knew this existed earlier, and I am already thrilled to know that it exists. I will be using this. This is actually pretty great. I like it. I like it a lot. Uh, very good, all right, rob you ready.
01:26:39 - Rob Campbell (Co-host)
Yes, okay so, as I was saying, inksy is a way I mean one of the pretty much what it's used for is to get system information. So a basic INXI-B is going to show you a bunch of your system information. For those who aren't seeing it, it's showing my system. It's the host, the kernel, the architecture, the desktop environment, machine hardware, cpu graphics, network drivers, lots of details on all those things. If I do a capital dash F, that does more in depth, so it says all the same things but more in depth. On that you can. Also, you know, if you do a dash capital B which I'm not on a laptop that's going to show you battery statistics. Since I am on a desktop I have no battery, so it's not showing me anything there. But there are. Also, if you don't want to see all that detail, you just want to see just the CPU detail, just the GPU, just the memory, etc. There are different flags for all of that. But what's really fun about this one is, even though it's a system information tool to quickly get information, it has one other really nice feature.
01:28:05
In this pre-show we were talking about temperatures, how hot it was in some places. Well, let me just show you how hot. It is where I am right now, at this very moment. So inxi, dash cap. No, actually I think it's lowercase w five, six, zero, zero, one, my zip code. You can all do it also do it like city, state, uh, country. For us in the us zip code is just an easy way to do it. There's like three different ways to do it, but anyway, if I hit that, it's going to show me the temperature here. Wow, it's really cooled off. It's apparently only 22.24 Celsius or 72 degrees Fahrenheit. Conditions are scattered clouds, and then it says locale on the time and the source is openweathermaporg. So you got a nice tool, incsee, to get system information and you know, while you're at it, for some reason you get the weather from it too.
01:29:09 - Jonathan Bennett (Host)
Nice. I can see why people use that to tell me about your system. Tell me the devices that are in it for doing how people do debug.
01:29:16 - Rob Campbell (Co-host)
Very nice. Yeah, it's a lot more detailed than NeoFetch.
01:29:20 - Jonathan Bennett (Host)
It's true, it just doesn't have the fancy graphics. Yeah, that's really what it's missing. All right, I've got a pair of. Well, there's sort of alternatives to Screen and TMUX, and pretty interesting ones. It's two projects, abduco and DVTM. So Abduco just lets you detach and retouch from a session, from a terminal, and that is all that it does. Like that's it. It just attaches and detaches. And then the DTVM all that it does is it's the dynamic manager, so that you can do tiling.
01:30:01
And so the idea here was someone said screen and Termux. They don't really use the Unix way. They're way too complicated, it's way too much computer code, and so these are very small, very simple programs. For example, dvtm is apparently only like 4,000 lines of C code, which is tiny compared to some of the other programs, and one of the other advantages actually about them is that their compiled version is really small, so you can put them in an embedded space. You could do an aptico and a DVTM with something like an OpenWrite router and have plenty of room to do it, whereas with Screen and Tmux it's just not going to fit. Looks like some pretty simple different it's taking a minute to learn it but some pretty simple keyboard shortcuts too. So if, for whatever reason, screen and or Tmux is not what you want, but you still want to be able to detach and reattach take a look at abdico and dvtm.
01:31:03 - Jeff Massie (Co-host)
It looks like. Yeah, what was it, jeff? I was gonna say?
01:31:05 - Jonathan Bennett (Host)
it looks like they haven't been updated since 2016. Are they?
01:31:07 - Jeff Massie (Co-host)
really that old. I mean one is in january and one it says march. It One is in January and one it says March. Based on the news.
01:31:13 - Jonathan Bennett (Host)
Yeah, you know what, it was March 24th of 2016. I saw the 24 and I thought, oh, it was updated just this year, and that's all my brain saw and all I thought. I think this is going to be one of those programs, though that it's simple enough. It's not going to need a whole bunch of updates, yeah.
01:31:34 - Rob Campbell (Co-host)
Well, it just works works.
01:31:34 - Jonathan Bennett (Host)
What do you need? It does? It does get the. Some programs, some real simple programs do get to the point, to where they're kind of done and you don't need to do a whole lot more to them need to add some features, like some fancy graphics like neo fetch I yeah well, you know, I think that's kind of out of scope, but you do, you man.
01:31:54 - Rob Campbell (Co-host)
Yeah, like the weather on Inksie.
01:31:56 - Jonathan Bennett (Host)
Yeah, like the weather, yes, All right, jeff, tell us about now.
01:32:08 - Jeff Massie (Co-host)
Is this O-Lama or is this O-Yama? I think it's O-Lama, o-l-l-a-m-a. I'm going with Olam. Okay, I'm going to talk about something a lot of people are sick of hearing, but some really want to play with AI there. I said it.
01:32:27
So Olam is an open source project that will let you run an LLM or a large language model locally. So there's several LLMs which you can get for free and some are even come pre-trained. The article linked in the show notes has a lot more, but a few that you can use are like Lama 3, and that's a family of open models which were developed by Meta GEMA G-E-M-M-A, which is an open model from Google DeepMind Minstrel, a model licensed under the Apache 2 license, and that's just a few of the ones you can run, and some are an entire suite of models. You are going to want a GPU to make them run at any reasonable speed and while you can run models in CPU only, they say even with a 16-core CPU, it'll just bring it to its knees and it's still going to run really slow. Olamo also doesn't integrated circuit for machine learning, and NPU is Neural Processing Unit, which is another special piece of hardware for AI types of workloads.
01:33:40
So what does Olam give you, rather than just running the LLM you know just natively? Probably the best way to describe it is, they say, think of Olam as a Docker container so you can run an LLM and then make sure it doesn't let any of the data you fed it out. Basically, it's just a nice little container, helps keep the LLMs organized and so that way you can run several different ones and they're not stepping on each other, not doing things you don't want them to do. So take a look at the article linked in the show notes for much more detail. They go into a lot of uh, under the hood stuff of how, how you can use it, and very, a lot more specifics. But you know, happy playing with ai yeah, very cool.
01:34:28 - Jonathan Bennett (Host)
Um, you know, there's some. There's a really fascinating things people have done with llms. One of the one of my favorites. There is speaking of mods, which I put mods early in the show. Uh, someone has worked on a couple of the, a couple of the famous video games and modded them to where the npc speech is now generated by llms and I thought that was uh, that was particularly interesting and big studios are taking taking of that, by the way. So, whether that's a good thing or a bad thing, I imagine that's coming. It's coming soon.
01:35:01 - David Ruggles (Co-host)
Quick side note about the Llama. If you play around with it, make sure you're on the latest version. There was a CVE for it last month. Well, it was publicly disclosed last month. It was actually properly handled, so it's several months old at this point with the remote code execution potential, but it's been fixed. So if you're on the latest version, you're good to go Go play with that AI.
01:35:28 - Jonathan Bennett (Host)
Yep, very cool. What's that, jeff? Oh, I was just agreeing. Yeah, okay, I thought you were trying to say something Get a point in before we close. It is that time, though, and so I will give each of the guys a chance to get in the last word or plug something, if they want to. We'll start with Rob, as we generally do. I think that's because Rob shows up first, and so he gets that coveted spot right next to me. That must be what it is, so, rob, take it away.
01:36:00 - Rob Campbell (Co-host)
So early bird gets the worm. Here I am. Come find my website at robertpcampbellcom. From there you can find links to my LinkedIn, my Twitter, my Mastodon place to buy me a coffee or you know what. If you even want to check out more of the site, you know you can scroll down and you can. You can find my resume and, you know, check if I'm somebody you want to hire or. Yeah, that's about all you can find about on me out there. But come connect, say hi, buy me a coffee, whatever's good.
01:36:37 - Jonathan Bennett (Host)
Yep, all right and Jeff.
01:36:42 - Jeff Massie (Co-host)
Nothing this week, so it's just a poem. There is a chasm of carbon and silicon. The software can't bridge. Have a good week everybody.
01:36:57 - David Ruggles (Co-host)
Excellent and David, the software camp bridge, have a good week everybody. Excellent. And david, how am I supposed to follow that? I don't know. So we talk about gaming off and on on here and, uh, I came across well, okay. First off, not everybody does it, but some people love trackballs for gaming. I am one of those people. I specifically love thumb trackballsballs and there is a company called game ball it's gaming trackballcom and they're doing some neat stuff and they're kind of trying some things out, kick-starting some stuff. So just check them out if game balls are something you would be interested in.
01:37:34 - Jonathan Bennett (Host)
Oh, I will have to look at that. I am also a trackball aficionado and I've I've started using and at least one, these huge game, the trackballs, and uh, I like those quite a bit. The problem I run into is that I wear the click the mouse buttons.
01:37:50 - David Ruggles (Co-host)
I wear the clickers out in the mouse buttons well, that's one of the things about this particular company. They've got the. Uh forget which ones they use, but it's like the cherry mx keyboard buttons, the high mechanical ones, that they've stuck in there good for a hundred thousand clicks, hopefully yeah, they've. They've upgraded a lot of the components.
01:38:12 - Jonathan Bennett (Host)
I will have to take a look at that because that does sound interesting. All right, thank you guys, and it is that time. We appreciate everybody being here. I do want to plug a couple of things real quick. Of course we've got Hackaday. You can find my work over there. We've got the security column. It goes live every Friday morning. That's where you follow to find things like the regression SSH flaw and all kinds of other stuff. So make sure to follow me there. We've also got Floss Weekly, which we have.
01:38:39
A bit of cross-pollination happens. So a couple of weeks ago we had the Amber language as one of our command line tips or maybe it was a story, but this past Wednesday actually on Tuesday, and then it went live on Wednesday we interviewed Paval, the creator of Amber, and that was a really fun conversation to have with him and get some insight into that language. So check that out as well. You can find that at hackadaycom. Slash floss and we should appreciate everybody being here, my co-hosts and everything. We appreciate those that watched live and those on the download, and we will see you next week on the Untitled Hank show. Hey, folks enjoying the show, but just wish you could do more or be more involved or catch it a little bit more live. Well, I've got news for you. There is club twit. It's about the price of a cup of coffee per month and you can be part of the club. Check it out, you don't want to miss it.
