Re: [PATCH RFC v4 07/44] KVM: guest_memfd: Only prepare folios for private pages

[Ackerley Tng <ackerleytng@google.com>]

Michael Roth <michael.roth@amd.com> writes:

> On Wed, Apr 01, 2026 at 07:05:16AM -0700, Ackerley Tng wrote:
>> Ackerley Tng <ackerleytng@google.com> writes:
>>
>> > All-shared guest_memfd used to be only supported for non-CoCo VMs where
>> > preparation doesn't apply. INIT_SHARED is about to be supported for
>> > non-CoCo VMs in a later patch in this series.
>> >
>> > In addition, KVM_SET_MEMORY_ATTRIBUTES2 is about to be supported in
>> > guest_memfd in a later patch in this series.
>> >
>> > This means that the kvm fault handler may now call kvm_gmem_get_pfn() on a
>> > shared folio for a CoCo VM where preparation applies.
>> >
>> > Add a check to make sure that preparation is only performed for private
>> > folios.
>> >
>> > Preparation will be undone on freeing (see kvm_gmem_free_folio()) and on
>> > conversion to shared.
>> >
>> > Signed-off-by: Ackerley Tng <ackerleytng@google.com>
>> > ---
>> >  virt/kvm/guest_memfd.c | 9 ++++++---
>> >  1 file changed, 6 insertions(+), 3 deletions(-)
>> >
>> > diff --git a/virt/kvm/guest_memfd.c b/virt/kvm/guest_memfd.c
>> > index b6ffa8734175d..d414ebfcb4c19 100644
>> > --- a/virt/kvm/guest_memfd.c
>> > +++ b/virt/kvm/guest_memfd.c
>> > @@ -900,6 +900,7 @@ int kvm_gmem_get_pfn(struct kvm *kvm, struct kvm_memory_slot *slot,
>> >  		     int *max_order)
>> >  {
>> >  	pgoff_t index = kvm_gmem_get_index(slot, gfn);
>> > +	struct inode *inode;
>> >  	struct folio *folio;
>> >  	int r = 0;
>> >
>> > @@ -907,7 +908,8 @@ int kvm_gmem_get_pfn(struct kvm *kvm, struct kvm_memory_slot *slot,
>> >  	if (!file)
>> >  		return -EFAULT;
>> >
>> > -	filemap_invalidate_lock_shared(file_inode(file)->i_mapping);
>> > +	inode = file_inode(file);
>> > +	filemap_invalidate_lock_shared(inode->i_mapping);
>> >
>> >  	folio = __kvm_gmem_get_pfn(file, slot, index, pfn, max_order);
>> >  	if (IS_ERR(folio)) {
>> > @@ -920,7 +922,8 @@ int kvm_gmem_get_pfn(struct kvm *kvm, struct kvm_memory_slot *slot,
>> >  		folio_mark_uptodate(folio);
>> >  	}
>> >
>> > -	r = kvm_gmem_prepare_folio(kvm, slot, gfn, folio);
>> > +	if (kvm_gmem_is_private_mem(inode, index))
>> > +		r = kvm_gmem_prepare_folio(kvm, slot, gfn, folio);
>>
>> Michael, I might have misunderstood you at the last guest_memfd call:
>> sev_gmem_prepare() doesn't prepare a page for being a shared page,
>> right? Does this work? That prepare is only called to "make private"?
>
> Hmm, I guess your guest_memfd-inplace-conversion-v4 branch is out of sync with
> these patches?
>

My bad, it was. I just force-pushed to github to synchronize them with
this patch series.

> I have the below local patch based on top of that for SNP-specific enablement,
> which is basically identically, so suffice to say: yes, this should work
> for SNP :) If any architecture pops up that needs to do some prep in
> advance of mapping shared pages, then we could potentially plumb the
> shared/private flag through to the arch-specific prep hook, as was also
> suggested on the call, but it doesn't seem like that's needed by any
> users for now.
>

Thanks for checking :)

> -Mike
>
>   Author: Michael Roth <michael.roth@amd.com>
>   Date:   Mon Oct 27 07:58:32 2025 -0500
>
>       KVM: guest_memfd: Don't prepare shared folios
>
>       In the current guest_memfd logic, "preparation" is only used currently
>       to describe the additional work of putting a guest_memfd page into an
>       architecturally-defined "private" state, such as updating RMP table
>       entries for SEV-SNP guests. As such, there's no input to the
>       corresponding kvm_arch_gmem_prepare() hooks as to whether a page is
>       being prepared/accessed as shared or as private, so "preparation" will
>       end up being erroneously done on pages that were supposed to remain in a
>       shared state. Rather than plumb through the additional information
>       needed to distinguish between shared vs. private preparation, just
>       continue to only do preparation on private pages, as was the case prior
>       to support for GUEST_MEMFD_FLAG_MMAP being introduced.
>
>       Signed-off-by: Michael Roth <michael.roth@amd.com>
>
>   diff --git a/virt/kvm/guest_memfd.c b/virt/kvm/guest_memfd.c
>   index 3acc6d983449..4869e59e4fc5 100644
>   --- a/virt/kvm/guest_memfd.c
>   +++ b/virt/kvm/guest_memfd.c
>   @@ -1249,7 +1249,8 @@ int kvm_gmem_get_pfn(struct kvm *kvm, struct kvm_memory_slot *slot,
>                   folio_mark_uptodate(folio);
>           }
>
>   -       r = kvm_gmem_prepare_folio(kvm, slot, gfn, folio);
>   +       if (!kvm_gmem_is_shared_mem(file_inode(file), index))
>   +               r = kvm_gmem_prepare_folio(kvm, slot, gfn, folio);
>
>           folio_unlock(folio);
>
>>
>> >
>> >  	folio_unlock(folio);
>> >
>> > @@ -930,7 +933,7 @@ int kvm_gmem_get_pfn(struct kvm *kvm, struct kvm_memory_slot *slot,
>> >  		folio_put(folio);
>> >
>> >  out:
>> > -	filemap_invalidate_unlock_shared(file_inode(file)->i_mapping);
>> > +	filemap_invalidate_unlock_shared(inode->i_mapping);
>> >  	return r;
>> >  }
>> >  EXPORT_SYMBOL_FOR_KVM_INTERNAL(kvm_gmem_get_pfn);
>> >
>> > --
>> > 2.53.0.1018.g2bb0e51243-goog