|
is there a reason to still have a separate boot partition? bioses stopped being dumb about booting from arbitrarily sized disk partitions in early 00s, and in efiland you need an efi partition but then the rest is arbitrary because efi's gone
|
# ? Nov 13, 2023 15:21 |
|
|
# ? May 24, 2024 05:46 |
Truga posted:is there a reason to still have a separate boot partition? bioses stopped being dumb about booting from arbitrarily sized disk partitions in early 00s, and in efiland you need an efi but then the rest is arbitrary Still, chainbooting is practiced more often than not.
|
|
# ? Nov 13, 2023 15:27 |
|
Klyith posted:Speaking of arch, is anyone using arch (or endeavour) having issues with system updates / package management related to the KDE 5->6 changeover? Like needing manual intervention when updating because Kdoodad got renamed to Kdoodad5, and some other K-package still wants the old name as a dependency. So instead of just pacman -Syu and walk away, you gotta look at things and force some changes. This happens with Arch maybe once or twice a year for me. Recently had it happen with JDK, too. You'll eventually find some jerk on the Arch forums explaining that it's obviously your fault for installing those packages which aren't in the official repos even though that's where they came from in the first place. Anyway, you can fix this by uninstalling the abandoned packages from the error message. Might have to do a few rounds of it, but KDE will eventually upgrade.
|
# ? Nov 13, 2023 15:31 |
|
Klyith posted:Speaking of arch, is anyone using arch (or endeavour) having issues with system updates / package management related to the KDE 5->6 changeover? Like needing manual intervention when updating because Kdoodad got renamed to Kdoodad5, and some other K-package still wants the old name as a dependency. So instead of just pacman -Syu and walk away, you gotta look at things and force some changes. I think that happens with all major kde updates on all rolling distros. I remember the same problem on gentoo with the kde 4-5 switch. Manjaro's strange bunching of upgrades will either make it better or worse, no idea which. And the fix will probably be hidden in an unrelated paragraph on their update blog.
|
# ? Nov 13, 2023 15:46 |
|
Truga posted:is there a reason to still have a separate boot partition? bioses stopped being dumb about booting from arbitrarily sized disk partitions in early 00s, and in efiland you need an efi partition but then the rest is arbitrary because efi's gone I was under the impression that when using btrfs, you need to put the kernel a separate boot partition with a more standard FS; fat, ext, etc. Is that not true? When I installed linux on this PC I did a manual partition/mounts setup, not a great choice when I didn't quite understand how it all worked. But I wanted the separate /home partition, and EXT4 as my root FS because someone ITT convinced me that btrfs was terrible. So I made a 300MB partition that was in fact just for EFI (so only about 1/2MB used). Later on when I realized that btrfs was not terrible and it would actually be super useful to have OS rollbacks, I tried to convert the system in-place to btrfs. I was following a hodge-podge of instructions that told me I needed to convert the EFI into a full /boot partition and move the kernel there, which I did successfully. The rest of the conversion was not successful, but at least the rollback worked perfectly. And I never bothered to undo the /boot change. I've been planing on switching distro and doing a btrfs root, in which case can I have just a tiny grub EFI partition and have the kernel in the main one? VictualSquid posted:I think that happens with all major kde updates on all rolling distros. I remember the same problem on gentoo with the kde 4-5 switch. Worse: the release thread in their forums. (Manjaro forums are pretty crap. I'd rather deal with hostility than incorrect info.) But at least they aren't uniquely to blame for the kde stuff.
|
# ? Nov 13, 2023 16:53 |
|
Truga posted:is there a reason to still have a separate boot partition? bioses stopped being dumb about booting from arbitrarily sized disk partitions in early 00s, and in efiland you need an efi partition but then the rest is arbitrary because efi's gone I think GRUB also has support for LUKS but back when I used LUKS I was unwilling to try it.
|
# ? Nov 13, 2023 16:55 |
|
BlankSystemDaemon posted:The MBR can only be 512 bytes long,
|
# ? Nov 13, 2023 16:55 |
|
Threw Ubuntu on a Surface Pro 6. God drat it is so much better than Windows. Had massive performance issues with the SP6 and recent Windows 11 versions, but Ubuntu is running like a dream with the surface linux kernel. Some of the performance issues may have come from running Windows Subsystem for Linux for dev purposes, which works like crap anyway so might as well go straight to the source Note for anyone who is thinking of throwing Ubuntu or another distro that supports secureboot on a Surface -- they just introduced a major irritation that only Mint Linux seems to handle gracefully. Really want to revisit dual booting on my desktop now.
|
# ? Nov 13, 2023 17:14 |
ExcessBLarg! posted:Even FreeBSD has supported native UEFI booting for like a decade now. Even if you do assume UEFI on everything, the ESP is a separate boot partition.
|
|
# ? Nov 13, 2023 17:17 |
|
Klyith posted:Speaking of arch, is anyone using arch (or endeavour) having issues with system updates / package management related to the KDE 5->6 changeover? Like needing manual intervention when updating because Kdoodad got renamed to Kdoodad5, and some other K-package still wants the old name as a dependency. So instead of just pacman -Syu and walk away, you gotta look at things and force some changes. I don't recall any of the Kdoodad → Kdoodad5 transitions being a problem, no. Only issue I've had recently was the aforementioned JDK - JRE thing: code:
|
# ? Nov 14, 2023 02:07 |
|
Vavrek posted:I don't recall any of the Kdoodad → Kdoodad5 transitions being a problem, no. Only issue I've had recently was the aforementioned JDK - JRE thing: You probably know this, but for the audience: A JDK is a Java Development Kit, and includes a java compiler, a bunch of other tools, and a JRE. A JRE is a Java Runtime Environment - everything you need to run java programs. In other words a JRE is a subset of a JDK. I can see how that could be fun to package.
|
# ? Nov 14, 2023 02:21 |
|
Computer viking posted:In other words a JRE is a subset of a JDK. I can see how that could be fun to package. Doesn’t seem that complicated, just make jdk depend on a matching jre and only include the extra stuff in the actual jdk package? I think it used to work like that on Red Hat back in the dark times.
|
# ? Nov 14, 2023 02:48 |
|
weird you guys aren't subscribed to the arch newsletter?? https://archlinux.org/news/incoming-changes-in-jdk-jre-21-packages-may-require-manual-intervention/
|
# ? Nov 14, 2023 06:33 |
|
Computer viking posted:You probably know this, but for the audience: I'm not sure why I had jdk-openjdk installed. I only know why I have jre-openjdk installed because of its Required By list. ziasquinn posted:weird you guys aren't subscribed to the arch newsletter??
|
# ? Nov 14, 2023 06:51 |
|
There is actually a pacman hook that will notify you of newsletter updates: https://github.com/bradford-smith94/informant If you have any aur packages they may also cause dependency issues if you forget to update them.
|
# ? Nov 14, 2023 17:23 |
|
M31 posted:There is actually a pacman hook that will notify you of newsletter updates: https://github.com/bradford-smith94/informant Thanks! That's exact the sort of thing I was thinking of. I think I saw that, or a reference to that, and it went in the back of my mind as an "oh yeah that sort of thing would be good probably I guess" and didn't resurface until the jdk situation.
|
# ? Nov 14, 2023 17:36 |
|
ziasquinn posted:weird you guys aren't subscribed to the arch newsletter?? I am, but in between the time of them posting the guide and me installing the updates, I forget about them. e: that hook sounds useless as it just makes you read the latest post irrespective of if you are installing the update that the post is about. Though it made me remember the portage hook that actually had logic linking the packages to the updates and showed you the update whenever you installed the package it explains. It pushed the info to page 79 of your 100 pages compiler output in the terminal, or if manually reconfigured into your syslog or system email. It also did so after installing the package in question, instead of before. VictualSquid fucked around with this message at 17:42 on Nov 14, 2023 |
# ? Nov 14, 2023 17:37 |
|
I'm using Mint + awesome wm. Extremely rarely, my system will crash and force a restart. When it does, I have an issue where I try to open Firefox and it will only open in my second workspace. The only way to fix it is to open an instance of Firefox in my main workspace (a 'bare' one without any of my pinned bookmarks, like SA) and then close the instance on the other workspace. Is there a way in Awesome to move windows from one workspace to another? It's not a huge deal; just a pain in the rear end and I'm sure there's an easier, more 'Linux-y' solution.
|
# ? Nov 14, 2023 22:35 |
|
What does the journal say? is this under x11? my gut is saying it's some kinda display environment variable getting owned by the crash?
|
# ? Nov 15, 2023 00:20 |
|
F_Shit_Fitzgerald posted:I'm using Mint + awesome wm. Extremely rarely, my system will crash and force a restart. When it does, I have an issue where I try to open Firefox and it will only open in my second workspace. The only way to fix it is to open an instance of Firefox in my main workspace (a 'bare' one without any of my pinned bookmarks, like SA) and then close the instance on the other workspace. Casual google says awesome wm calls workspaces "tags", so is that meta+shift+1 to put it on tag 1? F_Shit_Fitzgerald posted:I'm sure there's an easier, more 'Linux-y' solution. If you're using awesome wm I think the Linux-y solution is "you write your own lua script to control exactly how you want your windows to behave". Which is a very Linux-y solution. Dunno about easier. Especially if you don't know how to program in lua.
|
# ? Nov 15, 2023 02:00 |
|
Klyith posted:Casual google says awesome wm calls workspaces "tags", so is that meta+shift+1 to put it on tag 1? Perfect! Thanks. quote:If you're using awesome wm I think the Linux-y solution is "you write your own lua script to control exactly how you want your windows to behave". Yeah....I'm all about customization but I'm too lazy to learn how to write lua.
|
# ? Nov 15, 2023 02:12 |
|
Going to try and read the contents of your hard drive across hundreds of miles… is your /tmp mounted as tmpfs or in a real location
|
# ? Nov 15, 2023 02:54 |
|
ziasquinn posted:ofc the week I'm evangelizing Nixos I end up back on arch, my mother-distro What impelled you to switch back, if I may ask? I was reflecting that secondary to perfidious nvidia the thing that has wasted the most of my time on fedora i3 spin is figuring out exactly how they've implemented the defaults when I've needed to change them, and as I have things set up fairly minimally anyway it might suit me to have a barebones distro to set up so at least I know how it is setup, and arch and nixos are the top contenders. I might try to model the vanillaos way of using flatpaks / distrobox for everything which should solve the issue that I have a bit more stable releases and it might even be possible to have switchable DE's with this containerized approach even on arch.
|
# ? Nov 16, 2023 18:11 |
|
While lying in bed trying to stew this cold off I saw this article pop up on my phone's news recommendations and wanted to run it by this thread, although in the meantime it seems he has added some clarifications https://thehftguy.com/2023/11/14/the-linux-kernel-has-been-accidentally-hardcoded-to-a-maximum-of-8-cores-for-nearly-20-years/ I was curious if someone could say what would be the actual performance effects of this hard-locking to 8 cores, since I'd figure that you'd have to be doing something extraordinary to notice the difference on a standard desktop.
|
# ? Nov 16, 2023 21:43 |
|
Tesseraction posted:While lying in bed trying to stew this cold off I saw this article pop up on my phone's news recommendations and wanted to run it by this thread, although in the meantime it seems he has added some clarifications https://thehftguy.com/2023/11/14/the-linux-kernel-has-been-accidentally-hardcoded-to-a-maximum-of-8-cores-for-nearly-20-years/ It's a clickbait headline with some fundamental misunderstandings. What is capped to a maximum of 8 cores is a calculation on how long the time slices should be for the task scheduler, as in "what is the minimum amount of time a thread is allowed to run for before being preempted if the system is under load." That article is getting rebuttals all over the place but I don't know which rebuttal article is best yet. It was an intentional decision to cap that calculation at 8 cores, not an oversight. On low core count systems, this is low to improve responsiveness, and make sure that each thread is getting run regularly. On high core count systems the minimum time slice can be longer, but they saw negative effects even on high core count systems if you run a really long time slice, so they set the value calculated for 8 cores as the max.
|
# ? Nov 16, 2023 21:49 |
|
All right, I'll wait for the smoke to clear on the rebuttal articles before reading further, but it seemed like the kind of thing that someone would have raised before now if it was an actual problem, i.e. people saying "why is my kernel so loving slow?" and finding that value.
|
# ? Nov 16, 2023 21:52 |
|
It would be noticed by a hyperscaler in about 10 nanoseconds if the 64+ core systems they racked only used 8 of the cores.
|
# ? Nov 16, 2023 23:48 |
|
I mean, any HPC platform in the last 20 years, and any AMD user in the threadripper era, really.
|
# ? Nov 16, 2023 23:55 |
|
zhar posted:What impelled you to switch back, if I may ask? honestly I just wasn't fully utilizing the environment and knew that despite being rolling, it is always going to lag behind the one-and-holy arch. a little nagging feeling, i guess.
|
# ? Nov 17, 2023 00:03 |
|
But what if it only affects the kernel. What can the kernel do that would require more than 8 cores? Some heavy IO on bunch of NVMe drives? Kernel NFS and SMB daemons would probably be bottlenecked by the network first.
|
# ? Nov 17, 2023 00:07 |
|
How could it only affect the kernel? You mean that tasks making a syscall or triggering a fault would be rescheduled to one of the 8 cores if they were outside them in userspace? What happens to the processes that were already running there?
|
# ? Nov 17, 2023 00:11 |
|
Saukkis posted:But what if it only affects the kernel. What can the kernel do that would require more than 8 cores? Some heavy IO on bunch of NVMe drives? Kernel NFS and SMB daemons would probably be bottlenecked by the network first. This affects the scheduler, the kernel is the one choosing which userspace threads run and how.
|
# ? Nov 17, 2023 00:13 |
|
Subjunctive posted:How could it only affect the kernel? You mean that tasks making a syscall or triggering a fault would be rescheduled to one of the 8 cores if they were outside them in userspace? What happens to the processes that were already running there? In what kind of situation would the kernel need more than 800% of CPU power? That's how the title sounds the writer imagined the issue. Similar to how games can often effectively use only few cores, but they can still bounce around all the cores. I see all the time situations where 'top' shows a process using hundreds or thousands of % of CPU, but it's always python, R or some other random process, never kernel. I hardly ever see kernel threads using noticeable amount of CPU.
|
# ? Nov 17, 2023 00:23 |
zhar posted:What impelled you to switch back, if I may ask? If you decide to throw everything you're installing in a distrobox that can't be in a flatpak make sure to setup an auto-upgrade timer. Or if you build from source the distrobox git repo has a distrobox-assemble tool that's a little more robust than the current packaged version. You can define which bins or apps to export to the host to make it an entirely declarative config file and just have a systemd service set on a timer to rebuild your containers instead of updating them.
|
|
# ? Nov 17, 2023 00:23 |
|
Saukkis posted:In what kind of situation would the kernel need more than 800% of CPU power? That's how the title sounds the writer imagined the issue. Similar to how games can often effectively use only few cores, but they can still bounce around all the cores. Have you never seen those 64+ Python or R threads all fight over write access to the same memory? The syscalls for synchronization result in spending like 4000%+ of CPU time in the kernel, you get an htop full of red. You can do this super easily with R code that multiplies a lot of small matrices and the default OpenBLAS settings, which will spawn 1 thread per hardware thread and then have them all fight. It sucks. The OpenBLAS project gets GitHub issues raised about it all the time, and sheepishly say that the original development was done with zero regard for thread safety.
|
# ? Nov 17, 2023 00:29 |
|
ziasquinn posted:honestly I just wasn't fully utilizing the environment and knew that despite being rolling, it is always going to lag behind the one-and-holy arch. That's fair I think what has kept me off nixos is the feeling it's a little much for what I actually need Nitrousoxide posted:If you decide to throw everything you're installing in a distrobox that can't be in a flatpak make sure to setup an auto-upgrade timer. Thanks this is something I would have overlooked for sure
|
# ? Nov 17, 2023 00:30 |
|
Saukkis posted:In what kind of situation would the kernel need more than 800% of CPU power? That's how the title sounds the writer imagined the issue. Similar to how games can often effectively use only few cores, but they can still bounce around all the cores. But the kernel does work on behalf of userspace. When you make a syscall the kernel does a bunch of stuff in the context of that process. “kernel threads” are just processes (task_structs or similar) that don’t have a userspace context set up—most of the work that the kernel does is not by kswapd or kworker or whatever the current ones are. The kernel isn’t a separate process that has threads, like a normal program creates with pthread_whatever. It’s just a privileged memory mapping, so code loaded there can do all the ring 0 stuff and is protected from the ring 3 savages. when a process makes a system call, it’s calling the kernel like in would call a library, in the same thread of execution. It’s not messaging another process that’s accounted-for separately (though some monitoring tools do split CPU time between kernel and user contexts). So what could a kernel do to saturate 8 cores? If it weren’t for internal lock contention you could probably have the kernel suck up a lot of time if you slammed it with mapping changes or SYSV IPC or read/wrote in a tight loop from /sys.
|
# ? Nov 17, 2023 00:36 |
|
Tesseraction posted:All right, I'll wait for the smoke to clear on the rebuttal articles before reading further, These clickbait articles are the next level of that. Soon it will be YouTube videos, then YouTube rebuttals, and apologies, etc.
|
# ? Nov 17, 2023 02:03 |
|
Hopefully we get some kernel mailing list reaction videos
|
# ? Nov 17, 2023 02:07 |
|
|
# ? May 24, 2024 05:46 |
|
It happens in the security world too. Young bucks trying get their name out there make a mountain of a molehill. Sometimes it works, usually it doesn't. Depends on which news sites fall for it.
|
# ? Nov 17, 2023 02:11 |