Printk Changes Submitted For Linux 6.12 Finish NBCON Console Preparations

[phoronix]

Phoronix: Printk Changes Submitted For Linux 6.12 Finish NBCON Console Preparations

Ahead of the Linux 6.12 kernel merge window opening on Monday, the printk updates were submitted in advance given the Linux Kernel Maintainer Summit also taking place next week in Vienna. Notable with the printk updates is finishing up the NBCON console work that is notable as the last major blocker before real-time (PREEMPT_RT) support can be finally mainlined...

https://www.phoronix.com/news/Linux-6.12-Printk-Pull-Request

[varikonniemi]

This is some of the difficult low paying improvements that RT brings to make the kernel better for everyone.

[wertigon]

Big thank you to Thomas Gleixner, John Ogness and the rest of the team working on this and bringing it to mainline.

These guys deserve every scrap of recognition and applause the community can muster. Let me know if there are any donations I can encourage my employer to make.

[emansom]

If I understand correctly, the point of these patches is to lower the slice of time the kernel printk takes? And in that way the latency of the realtime patches can better be guaranteed?

[wertigon]

If I understand correctly, the point of these patches is to lower the slice of time the kernel printk takes? And in that way the latency of the realtime patches can better be guaranteed?

Kinda, but it's more like making printk preemptible so it will not hog the entire CPU at all times until it is done.

[siny]

Everything is pointless without scroll-back. Last 25, or even 50-60 lines are often not enough to see what went wrong...

[Volta]

Big thank you to Thomas Gleixner, John Ogness

Big thanks to Sebastian Andrzej Siewior and Petr Mladek.

[EphemeralEft]

How will “NBCON” be drawn on the screen? Will the drawing also be preemtible?

[gnarlin]

I fear I may be gaslighting myself. When did the preempt patches first appear? I feel like I saw them as a teenager, and I'm 44 years old!

[Developer12]

I'm really curious actually how rust kernel code or a rust kernel might approach this sort of Emergency/Panic takeover of a console. Honestly, it could go either way.

The really difficult requirement is to be able to take over the console no matter what state it's in when a panic occurs (not sure but it seems Emergency has to wait until it's safe when stealing from Normal?). On the one hand, rust has really great mechanics for handling ownership and lifetime. On the other hand, I don't know if it really supports "stealing" ownership of stuff. I guess a situation where the kernel is *already* actively in the process of panicking would be a fine place to use some unsafe code, but it doesn't really solve the problem. TBH I'm not even really sure exactly how the C code accomplishes it, the language where you can freely lie/cheat/steal.