[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] Xen Introspection, KPTI, and CR3 bit 63 leads to guest VMENTRY failures during introspection

On 01/26/2018 06:17 PM, Bitweasil . wrote:
> The proposed changes would only have an impact if CR3 exiting is
> enabled, which implies a pair of world switches and other code execution
> in a different region of memory and with different page tables anyway.
> Under normal operation, CR3 exiting is not turned on, so this will have
> no impact on operation.
> Are there any non-introspection cases in which CR3 exiting will be
> enabled for hardware virtualized guests?  Given the time cost of a pair
> of world switches and handling the associated code, I question if one
> could even measure the difference of the TLB flush or not.  The CR3
> reporting performance hit under KPTI is quite catastrophic anyway,
> though I expect one could mitigate that somewhat with CR3-targets, if
> one wanted to.

CR3 exits seem indeed to be enabled via {hvm,vmx}_update_cr(v, 0),
however I believe that they currently remain enabled even after the
vm_event consumer detaches from the guest (I dont think
CPU_BASED_CR3_LOAD_EXITING is being cleared from
v->arch.hvm_vmx.exec_control). This needs testing, but I believe it to
be correct.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.