intel CPU vulnerability NO_SSB whitelist for AIRMONT

intel CPU vulnerability NO_SSB whitelist for AIRMONT

[Shanavas.K.S]

Hi All,
This is in reference to
https://github.com/torvalds/linux/blame/master/arch/x86/kernel/cpu/common.c#L1169.
The whitelist for INTEL_ATOM_AIRMONT has NO_SSB enabled  which means
cache speculation vulnerability(spectre 4,CVE-2018-3639) is not
applicable to  this processor. But NO_SSB is not set for
INTEL_ATOM_AIRMONT_NP and INTEL_ATOM_AIRMONT_MID. Are these settings
missing for AIRMONT_NP and AIRMONT_MID or is there a real difference
between cache speculation between these processors?

regards
Shanavas

Re: intel CPU vulnerability NO_SSB whitelist for AIRMONT

[Pawan Gupta]

On Tue, Jul 09, 2024 at 01:21:44PM +0530, Shanavas.K.S wrote:
> Hi All,
> This is in reference to
> https://github.com/torvalds/linux/blame/master/arch/x86/kernel/cpu/common.c#L1169.
> The whitelist for INTEL_ATOM_AIRMONT has NO_SSB enabled  which means
> cache speculation vulnerability(spectre 4,CVE-2018-3639) is not
> applicable to  this processor. But NO_SSB is not set for
> INTEL_ATOM_AIRMONT_NP and INTEL_ATOM_AIRMONT_MID. Are these settings
> missing for AIRMONT_NP and AIRMONT_MID or is there a real difference
> between cache speculation between these processors?

I believe they were released much later than AIRMONT and hence should be
enumerating MSR IA32_ARCH_CAPABILITIES[SSB_NO](bit 4). If you have these
system, is bit 4 not set?

If it is set there is no need to explicitly add NO_SSB to the table.

Re: intel CPU vulnerability NO_SSB whitelist for AIRMONT

[Shanavas.K.S]

 Thanks for the response.  Please create the patch from your side

Regards
Shanavas
> On Thu, Aug 8, 2024 at 12:32 AM Pawan Gupta
> <pawan.kumar.gupta@linux.intel.com> wrote:
> >
> > On Wed, Aug 07, 2024 at 12:57:47PM +0530, Shanavas.K.S wrote:
> > > Hi Pawan Gupta
> > > Thanks for the comment. I have checked  IA32_ARCH_CAPABILITIES
> > > register but it is resulting in "unchecked MSR access error: RDMSR
> > > from 0x10a at rIP: 0xffffffff8286cc26". I am assuming the reason is
> > > the  MSR register is not available in our processor.
> >
> > Thats what it looks like.
> >
> > > Does it mean that NO_SSB should be added to the whitelist for
> > > INTEL_ATOM_AIRMONT_NP and INTEL_ATOM_AIRMONT_MID if these processors are
> > > not susceptible to cache speculation vulnerability?
> >
> > Correct, NO_SSB should be added to the whitelist for AIRMONT variants. Let
> > me know if you are planning to submit a patch for this. Otherwise, I will.