Linux Will Be Able To Boot ~0.035 Seconds Faster With One Line Kernel Patch

[phoronix]

Phoronix: Linux Will Be Able To Boot ~0.035 Seconds Faster With One Line Kernel Patch

The Linux kernel itself can already boot quite fast but with a simple one-line patch another ~0.035 seconds will be able to be shaved off the boot time...

https://www.phoronix.com/news/Linux-Faster-Boot-One-Line-ACPI

[geerge]

Align all of the things

[emansom]

Gonna call it now: some university student is gonna point out how this allows for a boot attack. Four years from now or something.

[rabcor]

booting faster isn't really the problem, it's shutting down that is the problem.

I mean we have watchdog shenanigans preventing shutdowns, we have systemd shenanigans preventing shutdown, and we have driver crashes that freeze the whole kernel up (the permanent way) on shutdown, booting? booting is never the damn problem. We've been booting faster than windows for over a decade.

[imipak]

If that mattered, data centres would use Coreboot and run level 3. Coreboot's website suggests that this can give a boot time of 400ms, but that's presumably not to a full system. A full system startup should, nonetheless, be very quick.

Data centres would also not be using the default kernel, but rather a build tuned to the system.

The reality is that I've never known a data centre to do either of these things, even though it would shorten boot times significantly. So boot times, although important, are clearly not critical.

[LtdJorge]

booting faster isn't really the problem, it's shutting down that is the problem.

I mean we have watchdog shenanigans preventing shutdowns, we have systemd shenanigans preventing shutdown, and we have driver crashes that freeze the whole kernel up (the permanent way) on shutdown, booting? booting is never the damn problem. We've been booting faster than windows for over a decade.

Hmm, my system shutsdown in like 3 seconds max. If I have a VM running with the guest agent installed, it may take 5.

It has only taken more when I have broken something like a mount/device and systemd waits 2 minutes for unmounting.

[LinAdmin]

What a fantastic, overwhelming improvement.
All my thanks to phoronix for informing me about every peanut :P

[rabcor]

What a fantastic, overwhelming improvement.
All my thanks to phoronix for informing me about every peanut :P

hey now, 35ms is almost enough for you to feel it if you're paying very close attention! almost!

Hmm, my system shutsdown in like 3 seconds max. If I have a VM running with the guest agent installed, it may take 5.

It has only taken more when I have broken something like a mount/device and systemd waits 2 minutes for unmounting.

That's great, and you never had to do anything special to make sure it does that?

I've had boot hangs due to systemd or watchdog that lead to me needing to use sysrq commands to finish shutdown if i want it to happen at all (like boot loops, but shutdown loops instead!) and I also semi frequently encounter a (known) issue where when the nvidia drivers are turned off the entire system just freezes. Now you might think "oh, nvidia, of course, nvidia bad, nvidia's fault" but no, the kernel freezing up is firmly on the kernel, it doesn't matter how bad the nvidia driver is, the kernel should be able to handle it at least to such an extent that it can complete the shutdown process, because it should have safeguards to prevent this kind of thing from happening.

[Kver]

All the time this will save humanity has been wasted in this thread.

[alpha_one_x86]

speaking of boot time, most of the server I have boot the BIOS/UEFI into 3-7min! Then to improve boot time on this kind of server/hyperscaler, maybe improve BIOS/UEFI boot time will help