From: Borislav Petkov <bp@xxxxxxxxx> Sent: Thursday, December 29, 2022 4:18 AM
>
> On Thu, Dec 01, 2022 at 07:30:22PM -0800, Michael Kelley wrote:
> > Current code in sme_postprocess_startup() decrypts the bss_decrypted
> > section when sme_me_mask is non-zero. But code in
> > mem_encrypt_free_decrypted_mem() re-encrypts the unused portion based
> > on CC_ATTR_MEM_ENCRYPT. In a Hyper-V guest VM using vTOM, these
> > conditions are not equivalent as sme_me_mask is always zero when
> > using vTOM. Consequently, mem_encrypt_free_decrypted_mem() attempts
> > to re-encrypt memory that was never decrypted.
> >
> > Fix this in mem_encrypt_free_decrypted_mem() by conditioning the
> > re-encryption on the same test for non-zero sme_me_mask. Hyper-V
> > guests using vTOM don't need the bss_decrypted section to be
> > decrypted, so skipping the decryption/re-encryption doesn't cause
> > a problem.
>
> Lemme simplify the formulations a bit:
>
> "sme_postprocess_startup() decrypts the bss_decrypted ection when me_mask
> sme_is non-zero.
>
> mem_encrypt_free_decrypted_mem() re-encrypts the unused portion based on
> CC_ATTR_MEM_ENCRYPT.
>
> In a Hyper-V guest VM using vTOM, these conditions are not equivalent
> as sme_me_mask is always zero when using vTOM. Consequently,
> mem_encrypt_free_decrypted_mem() attempts to re-encrypt memory that was
> never decrypted.
>
> So check sme_me_mask in mem_encrypt_free_decrypted_mem() too.
>
> Hyper-V guests using vTOM don't need the bss_decrypted section to be
> decrypted, so skipping the decryption/re-encryption doesn't cause a
> problem."
Work for me. I'll pick up the new wording in v5.
>
> > Fixes: e9d1d2bb75b2 ("treewide: Replace the use of mem_encrypt_active() with
> cc_platform_has()")
>
> So when you say Fixes - this is an issue only for vTOM-using VMs and
> before yours, there were none. And yours needs more enablement than just
> this patch.
>
> So does this one really need to be backported to stable@?
>
> I'm asking because there's AI which will pick it up based on this Fixes
> tag up but that AI is still not that smart to replace us all. :-)
>
I'm ambivalent on the backport to stable. One might argue that older
kernel versions are conceptually wrong in using different conditions for
the decryption and re-encryption. But as you said, they aren't broken
from a practical standpoint because sme_me_mask and
CC_ATTR_MEM_ENCRYPT are equivalent prior to my patch set. However,
the email thread with Sathyanarayanan Kuppuswamy, Tom Lendacky,
and Dexuan Cui concluded that a Fixes: tag is appropriate. See
https://lore.kernel.org/lkml/fbf2cdcc-4ff7-b466-a6af-7a147f3bc94d@xxxxxxx/
and
https://lore.kernel.org/lkml/BYAPR21MB1688A31ED795ED1B5ACB6D26D7099@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/
Michael
