Re: [PATCH v8 2/2] x86/mm: Don't disable PCID if the kernel is running on a hypervisor

From: Xi Ruoyao
Date: Wed Apr 17 2024 - 04:02:51 EST


On Tue, 2024-04-16 at 16:49 -0700, Sean Christopherson wrote:
> On Sat, Apr 13, 2024, Xi Ruoyao wrote:
> > The Intel erratum for "incomplete Global INVLPG flushes" says:
> >
> >     This erratum does not apply in VMX non-root operation. It applies
> >     only when PCIDs are enabled and either in VMX root operation or
> >     outside VMX operation.
> >
> > So if the kernel is running in a hypervisor, we are in VMX non-root
> > operation and we should be safe to use PCID.
> >
> > Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
> > Cc: Michael Kelley <mhklinux@xxxxxxxxxxx>
> > Cc: Pawan Gupta <pawan.kumar.gupta@xxxxxxxxxxxxxxx>
> > Cc: Sean Christopherson <seanjc@xxxxxxxxxx>
> > Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> > Link: https://lore.kernel.org/all/168436059559.404.13934972543631851306tip-bot2@tip-bot2/
> > Link: https://cdrdv2.intel.com/v1/dl/getContent/740518 # RPL042, rev. 13
> > Link: https://cdrdv2.intel.com/v1/dl/getContent/682436 # ADL063, rev. 24
> > Signed-off-by: Xi Ruoyao <xry111@xxxxxxxxxxx>
> > ---
> >  arch/x86/mm/init.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> >
> > diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
> > index c318cdc35467..6010f86c5acd 100644
> > --- a/arch/x86/mm/init.c
> > +++ b/arch/x86/mm/init.c
> > @@ -275,6 +275,14 @@ static void __init probe_page_size_mask(void)
> >   * microcode is not updated to fix the issue.
> >   */
> >  static const struct x86_cpu_id invlpg_miss_ids[] = {
> > + /* Only bare-metal is affected.  PCIDs in guests are OK.  */
> > + {
> > +   .vendor = X86_VENDOR_INTEL,
> > +   .family = 6,
> > +   .model = INTEL_FAM6_ANY,
> > +   .feature = X86_FEATURE_HYPERVISOR,
>
> Isn't this inverted?  x86_match_cpu() will return NULL if the CPU doesn't have
> HYPERVISOR.  We want it to return NULL if the CPU *does* have HYPERVISOR.

Hmm, but it seems not possible to let x86_match_cpu() to always return
NULL if the CPU does have HYPERVISOR. If I read x86_match_cpu()
correctly it cannot do an inverted feature match. Or am I
misunderstanding something here?

Instead this makes x86_match_cpu() return an entry with driver_data = 0
if the CPU have HYPERVISOR, thus boot_cpu_data.microcode <
invlpg_miss_match->driver_data will always be false when the CPU have
HYPERVISOR and PCID won't be disabled.

> > +   .driver_data = 0,
> > + },
> >   INTEL_MATCH(INTEL_FAM6_ALDERLAKE, 0x2e),
> >   INTEL_MATCH(INTEL_FAM6_ALDERLAKE_L, 0x42c),
> >   INTEL_MATCH(INTEL_FAM6_ATOM_GRACEMONT, 0x11),
> > --
> > 2.44.0
> >

--
Xi Ruoyao <xry111@xxxxxxxxxxx>
School of Aerospace Science and Technology, Xidian University