Re: Yet another KPTI regression with 4.14.x series in a VM

[Peter Zijlstra <peterz@infradead.org>]

On Fri, Jan 12, 2018 at 10:08:20PM -0800, Andy Lutomirski wrote:
> Now this is quite a strange value to write to CR3.  The 0x800 part
> means that we're using the "user" variant of the address space that
> would have ASID=0 and the 0x1000 bit being set corresponds to the user
> pgdir, but this is nonsense, since the kernel never uses PCID 0 for
> user mode.  We always start at 1.  The only exception is if
> X86_FEATURE_PCID is off.  But, if X86_FEATURE_PCID is off, then we
> shouldn't be setting any PCID bits.

My bad, I was under the impression the lower 12 bits would be ignored
without PCID :/

> .Lwrcr3_\@:
>         /* Flip the PGD and ASID to the user version */
>         orq     $(PTI_SWITCH_MASK), \scratch_reg
>         mov     \scratch_reg, %cr3
> .Lend_\@:
> 
> That's bogus.  PTI_SWITCH_MASK is 0x1800, which has PCID = 0x800.

> This should probably use an alternative to select between 0x1000 and
> 0x800 depending on X86_FEATURE_PCID or just use an entirely different
> label for the !PCID case.

	ALTERNATIVE "orq $(PTI_SWITCH_PGTABLE_MASK), \scratch_reg",
	            "orq $(PTI_SWITCH_MASK), \scratch_reg", X86_FEATURE_PCID

Is not wanting to compile though; probably that whole alternative vs
macro thing again :/