Linux 5.3 Could Finally See FSGSBASE - Performance Improvements Back To Ivybridge

[phoronix]

Phoronix: Linux 5.3 Could Finally See FSGSBASE - Performance Improvements Back To Ivybridge

The FSGSBASE instruction set has been present on Intel processors going back to Ivy Bridge processors and while there have been Linux kernel patches for this feature going on for years, it looks like with the Linux 5.3 kernel cycle is this support for merging. Making us eager for this support is the prospect of better performance, especially for context switching workloads that already have been suffering as a result of recent CPU mitigations...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-5.3-FSGSBASE

[tildearrow]

Now userspace is going to do totally stupid shite like writing some nonzero value to GS and then doing WRGSBASE or like linking some idiotic library that uses WRGSBASE into a perfectly innocent program like dosemu2 and resulting in utterly nonsensical descriptor state.

I really miss this good old Torvalds...

[Guest]

I hope these instructions won't cause more headaches with vulnerabilities.

[PCJohn]

I really miss this good old Torvalds...

Me not at all. I recommend to not connect word "good" and offensive language. By the way, offensive language might be connected with unbalanced personality.... Not connecting this with anybody! and note word "might". Just mentioning it and rather considering such language as negative.

[treba]

I hope these instructions won't cause more headaches with vulnerabilities.

Exactly what I thought, too :/ But given how much attention is given to that know I'd trust the devs to have considered that thrououtly.

[stormcrow]

Among the reasons the code has been delayed in previous years is that user-space can do stupid stuff, "The major disadvantage is that user code can use the new instructions. Now userspace is going to do totally stupid shite like writing some nonzero value to GS and then doing WRGSBASE or like linking some idiotic library that uses WRGSBASE into a perfectly innocent program like dosemu2 and resulting in utterly nonsensical descriptor state."

I don't buy this explanation for not implementing a CPU feature. If we used that fundamental argument every time something new comes out *nothing* would ever be implemented because there's quite literally nothing in hardware that can't be abused by an inexperienced or even an experienced developer. Having a network stack with raw packet crafting support can be abused in various nefarious ways, but Linux has that, too.

[c2p_]

I really miss this good old Torvalds...

Torvalds isn't an author of these words but Andy Lutomirski. Old Linus is still present on mailing lists, nothing changed.