Linux 6.9 Sees Invasive & Significant Changes To Workqueues

[phoronix]

Phoronix: Linux 6.9 Sees Invasive & Significant Changes To Workqueues

Workqueues are commonly used within the Linux kernel for asynchronous process execution contexts. With Linux 6.9 the workqueue (WQ) code has seen "significant and invasive" changes...

https://www.phoronix.com/news/Linux-6.9-Workqueue

[Mitch]

I can't find the PR or referenced documentation from the article. I'd love to read about it

[Michael]

I can't find the PR or referenced documentation from the article. I'd love to read about it

Whoops, accidentally forgot the PR link, it's in there now. Thanks.

[sophisticles]

When i read these types of things i can't help but think that they need to adopt a "if it ain't broke, don't fix it" policy, after they get done fixing all the broken things of course.

At this point the kernel development should be nothing more than adding new drivers to support new hardware; no more trying to add some new "features" that end up breaking something they don't discover for years, not let's see if we can tweak this or clean up the code, that should be taken care of before the commit is made, just a new piece of hardware is out, let's add support for it.

This seems to be a mentality with a lot of open source projects, their maintainers can never call the job done, they keep wanting to add stuff but in many cases they can't even fix the stuff they already have.

[coder]

When i read these types of things i can't help but think that they need to adopt a "if it ain't broke, don't fix it" policy, after they get done fixing all the broken things of course.

There are a few reasons for changes like these:

1. Improving resource utilization on existing computers.
2. Improving scalability for upcoming machines.
3. Improving usability or applicability of the infrastructure.
4. Easing maintenance of the code, which might be getting cluttered up, due to numerous rounds of optimizations oriented towards pet use cases or specific systems.

If you had your way, we wouldn't see improvements like this:

https://www.phoronix.com/review/ampere-altra-ubuntu-2404

This seems to be a mentality with a lot of open source projects, their maintainers can never call the job done, they keep wanting to add stuff but in many cases they can't even fix the stuff they already have.

Hardware is always changing. Software is evolving to utilize new hardware and address new customer demands. New security vulnerabilities & techniques are being discovered. The kernel cannot remain static, because it doesn't exist in a static world.

[Volta]

When i read these types of things i can't help but think that they need to adopt a "if it ain't broke, don't fix it" policy, after they get done fixing all the broken things of course.

Tell this to your m$ toy company:

https://www.windowslatest.com/2024/03/16/windows-11-kb5035853-march-2024-update-causes-major-issues-bsods/

“Can it really get worse from here?” That was my first thought after failing to install the January and February 2024 cumulative updates despite multiple attempts. Then, on the 13th, Microsoft released Windows 11 KB5035853. However, it seems to cause Blue Screen of Death errors for some users, including me. The only workaround is to remove the update from Settings > Updates > Uninstall updates.

This might sound like a neat Windows Update, but it’s not. After successfully installing the cumulative update, one of my devices experienced a Blue Screen of Death immediately after login. As shown in the below photo, the reason for BSOD is “Thread Stuck in Device Driver”, which doesn’t explain the problem.

In a Reddit thread spotted by Windows Latest, some users report their systems are “chugging hard,” with persistent stuttering and audio glitches, among other problems.

There’s also been a rise in the dreaded Blue Screen of Death (BSOD) errors since the update, with some mentioning Memory Management errors. Another big headache is that some users find they can’t start Windows 11 at all after the update.

One detailed complaint talked about blue screen errors saying “DPC_WATCHDOG_VIOLATION,” which could mean something’s wrong with the CPU or how the device manages power.

Here’s a list of all the issues we’ve spotted so far:

• Problems with system stuttering, hitching, and taking a long time to boot up.
• Problems with Lenovo L13 Ryzen Pro 7 machines, including boot loops and Bitlocker code requests. One user said: We are having issues with Lenovo L13 with Ryzen Pro 7… machines are going into a loop requesting the Bitlocker code.”
• Reports of blue screens with different error codes.

These issues mostly affect Windows 11 users, but at least one user with Windows 10 reached out to me to highlight their struggle with printers after KB5035845.

[cend]

LWN has more on this:

https://lwn.net/Articles/960041/

When i read these types of things i can't help but think that they need to adopt a "if it ain't broke, don't fix it" policy, after they get done fixing all the broken things of course.

If anything, tasklets is one of "all the broken things", because it stands in the way for "adding new drivers". Developers are yearning for a less broken, more efficient and intuitive API for the job done by tasklets for years.

我去，七十三老登，这状态怕不是写 COBOL 写的

[sophisticles]

Tell this to your m$ toy company:

https://www.windowslatest.com/2024/03/16/windows-11-kb5035853-march-2024-update-causes-major-issues-bsods/

In all fairness, i have stated numerous times that i do not believe in upgrading or updating Windows until i have no choice.

I said i stuck with Win 2k until I bought a Geforce 3 Ti4600 and needed DX8 for new games that I was playing.

I stuck with XP 32-bit until i bought an Athlon64 and needed a 64-bit OS.

I stuck with XP64 until I needed to upgrade to Win 7 and I stuck with Win 7 until I needed to upgrade to Win 10.

I recently upgraded to Win 11 and frankly there are some annoyances that makes me seriously consider wiping the drive clean and going back to Win 10.

For Win 10/11 I don't install any updates except for drivers and the H1/H2 updates after they are fully tested in the wild.

I don't update any of my Linux boxes unless i absolutely have to.

I have never believed in the update for the sake of update because a lot of times there are nasty bugs that aren't discovered until they are in the wild.

[cend]

I have never believed in the update for the sake of update because a lot of times there are nasty bugs that aren't discovered until they are in the wild.

If new software has nasty bugs in the wild, same for old software. Good luck fighting computer viruses.

[sophisticles]

LWN has more on this:

https://lwn.net/Articles/960041/

If anything, tasklets is one of "all the broken things", because it stands in the way for "adding new drivers". Developers are yearning for a less broken, more efficient and intuitive API for the job done by tasklets for years.

我去，七十三老登，这状态怕不是写 COBOL 写的

Maybe it's time to rethink the *nix architecture.

Linus released the early version of Linux in 1991 and by then Unix was 21 years old.

Today the basic *nix architecture has been around for over 50 years.

This may seem to contradict my "if it ain't broke, don't fix it", but it doesn't because I'm saying maybe it's time for a clean sheet design, maybe where the drivers all run in user mode.

I can't believe i am going to say this, but maybe start a new kernel, written in Rust and assembler, devoid of all the cruft that has accumulated over the years, with no backwards comparability, just say this kernel will only work with hardware released in 2025 and newer.

It just seems that at the current state of about 25 million lines of code in the current kernel, it has become unwieldy and we keep getting a steady stream of articles like this that make me wonder what that 177 million a year the Linux Foundation raises is going to.