Linux 5.13 Lands More Fixes To The Mucked Up FPU/XSTATE Handling Mess

[phoronix]

Phoronix: Linux 5.13 Lands More Fixes To The Mucked Up FPU/XSTATE Handling Mess

Earlier this month Linux 5.13 disabled Intel's ENQCMD functionality for upcoming Xeon "Sapphire Rapids" processors as the kernel software code around it was deemed "broken beyond repair". More of the recent Intel-submitted patches around reworking kernel code in preparation for upcoming CPU features has been found to be rather hairy after already being mainlined and thus another batch of urgent x86 fixes were sent in this morning...

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.13-More-XSTATE-Mess

[skeevy420]

We'll see what the "a lot more in the pipe" amounts to in the coming days

Seems to me that "a lot more in the pipe" is what got them into this mess.

[Snaipersky]

Unit tests Intel, what are they? Change the code, test the code. Do not submit it if thing in question breaks horribly.

[Sonadow]

Current workstation uses dual Xeon E5-2697 v2 processors with 128GB memory.

Not sure whether this affects me.

[misGnomer]

Whodunnit?

[Radtraveller]

Seems to me that "a lot more in the pipe" is what got them into this mess.

Do they all work in Oregon?

[Snaipersky]

Lol, unit tests for kernel code and hardware are tough. Maybe with an emulator that emulates the specific hardware in question.......

I should hope you have either an accurate emulator or actual hardware to test on, instead of submitting to mainline code that you hope works and then later ascertaining through unspecified means that it's absolute garbage and shouldn't have been committed in the first place.

[mdedetrich]

I should hope you have either an accurate emulator or actual hardware to test on, instead of submitting to mainline code that you hope works and then later ascertaining through unspecified means that it's absolute garbage and shouldn't have been committed in the first place.

Well since we are talking about x86 CPU's here, "accurately emulating" such a thing is close to impossible and if we are talking about unreleased hardware, the only company that has the hardware in question is Intel and they are the ones submitting the patch in the first place.

[Snaipersky]

Well since we are talking about x86 CPU's here, "accurately emulating" such a thing is close to impossible and if we are talking about unreleased hardware, the only company that has the hardware in question is Intel and they are the ones submitting the patch in the first place.

That's the point. If you look further up, I've been directing my criticism towards Intel, for submitting to mainline code that they themselves have not tested, either against their product or an emulation thereof.

[old_skull]

You make it all sound pretty straight forward, Snaipersky. Maybe they ought to hire you do mocking required. I am sure it'll be great