On 01/17/2018 09:04 AM, Andrew Cooper wrote:

On 17/01/18 09:02, Joerg Roedel wrote:

quoted

Hi Boris,

thanks for testing this :)

On Tue, Jan 16, 2018 at 09:47:06PM -0500, Boris Ostrovsky wrote:

quoted

On 01/16/2018 11:36 AM, Joerg Roedel wrote:

quoted

+.macro SWITCH_TO_KERNEL_STACK nr_regs=0 check_user=0

This (and next patch's SWITCH_TO_ENTRY_STACK) need X86_FEATURE_PTI check.

With those macros fixed I was able to boot 32-bit Xen PV guest.

Hmm, on bare metal the stack switch happens regardless of the
X86_FEATURE_PTI feature being set, because we always program tss.sp0
with the systenter stack. How is the kernel entry stack setup on xen-pv?
I think something is missing there instead.

There is one single stack registered with Xen, on which you get a normal
exception frame in all cases, even via the registered (virtual)
syscall/sysenter/failsafe handlers.

And so the check should be at least against X86_FEATURE_XENPV, not
necessarily X86_FEATURE_PTI.

But I guess you can still check against X86_FEATURE_PTI since without it
there is not much reason to switch stacks?

-boris

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>