Wine Takes Minor Performance Hit Running Windows Programs On Linux With KPTI

[Michael]

I have a question. Say I would like to upgrade my kernel that includes this patch but would like to turn off KPTI to avoid the performance loss, will I be able to do that and if so, can anyone please tell me how? Thanks.

I play a few games on Wine and Guild Wars 2 is one of the main reasons, which happens to be very CPU intensive, this might kill its' not-too-shabby performance.

As mentioned in the other articles on Phoronix this week, you can boot your kernel with "nopti" and verify via looking for the "page tables" message in dmesg.

[FireBurn]

Would be interesting to see if wine sees more of a hit than windows itself or vice versa

[jojo7887]

Thanks Michael, just the answer I was looking for :-)

[AJenbo]

Michael nice job testing this.

[schmidtbag]

That's lovely. That's like saying that because Microsoft wasn't vulnerable to the Heartbleed bug that input validation on Microsoft products is totally unnecessary. Which it totally stupid but is pretty much the logical position of AMD at this point.

I don't think you understand how this problem works... It's hardware specific. There are already kernel overrides for AMD.

[Dehir]

This is one of the more interesting areas to be tested. Other is running windows as virtualized qemu with vfio gpu passthrough.

[Dukenukemx]

From what I figure, the "fix" seems to put a load on the CPU. In Windows the games don't use as much of the CPU cause in general games don't use all the CPU. In Linux since a lot of stuff to run Windows games generally uses more CPU load, you'll notice it more. Probably less with Gallium-Nine.

Willing to bet Pentium and Core i3's will see a massive hit in performance, while quad cores and higher won't.

[nuetzel]

You 'all' have to grep for 'X86_BUG_CPU_MELTDOWN' in latest kernels.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=de791821c295cc61419a06fe5562288417d1bc58

[dungeon]

You 'all' have to grep for 'X86_BUG_CPU_MELTDOWN' in latest kernels.
https://git.kernel.org/pub/scm/linux...62288417d1bc58

They should rename both again to something like KAISERI MELTDOWNSKY

Or even better intel_inside_since_1995

[lucrus]

That's lovely. That's like saying that because Microsoft wasn't vulnerable to the Heartbleed bug that input validation on Microsoft products is totally unnecessary. Which it totally stupid but is pretty much the logical position of AMD at this point.

By the same reasoning, we should patch every OS out there to make it execute every process twice and then compare the results for security, because you never know, maybe your trusty ECC RAM has a bug or will have a bug in the future.

Please don't let Intel f*ck your brain with PR blurbs.