Hi Fuad!
On Wed, 2025-01-29 at 17:23 +0000, Fuad Tabba wrote:
> Add support for mmap() and fault() for guest_memfd backed memory
> in the host for VMs that support in-place conversion between
> shared and private (shared memory). To that end, this patch adds
> the ability to check whether the VM type has that support, and
> only allows mapping its memory if that's the case.
>
> Additionally, this behavior is gated with a new configuration
> option, CONFIG_KVM_GMEM_SHARED_MEM.
>
> Signed-off-by: Fuad Tabba <tabba@xxxxxxxxxx>
>
> ---
>
> This patch series will allow shared memory support for software
> VMs in x86. It will also introduce a similar VM type for arm64
> and allow shared memory support for that. In the future, pKVM
> will also support shared memory.
> ---
> include/linux/kvm_host.h | 11 ++++++
> virt/kvm/Kconfig | 4 +++
> virt/kvm/guest_memfd.c | 77 ++++++++++++++++++++++++++++++++++++++++
> 3 files changed, 92 insertions(+)
>
> -snip-
>
> diff --git a/virt/kvm/guest_memfd.c b/virt/kvm/guest_memfd.c
> index 47a9f68f7b24..86441581c9ae 100644
> --- a/virt/kvm/guest_memfd.c
> +++ b/virt/kvm/guest_memfd.c
> @@ -307,7 +307,84 @@ static pgoff_t kvm_gmem_get_index(struct kvm_memory_slot *slot, gfn_t gfn)
> return gfn - slot->base_gfn + slot->gmem.pgoff;
> }
>
> +#ifdef CONFIG_KVM_GMEM_SHARED_MEM
> +static vm_fault_t kvm_gmem_fault(struct vm_fault *vmf)
> +{
> + struct inode *inode = file_inode(vmf->vma->vm_file);
> + struct folio *folio;
> + vm_fault_t ret = VM_FAULT_LOCKED;
> +
> + filemap_invalidate_lock_shared(inode->i_mapping);
> +
> + folio = kvm_gmem_get_folio(inode, vmf->pgoff);
> + if (IS_ERR(folio)) {
> + ret = VM_FAULT_SIGBUS;
> + goto out_filemap;
> + }
> +
> + if (folio_test_hwpoison(folio)) {
> + ret = VM_FAULT_HWPOISON;
> + goto out_folio;
> + }
> +
> + if (WARN_ON_ONCE(folio_test_guestmem(folio))) {
> + ret = VM_FAULT_SIGBUS;
> + goto out_folio;
> + }
> +
> + /* No support for huge pages. */
> + if (WARN_ON_ONCE(folio_nr_pages(folio) > 1)) {
> + ret = VM_FAULT_SIGBUS;
> + goto out_folio;
> + }
> +
> + if (!folio_test_uptodate(folio)) {
> + clear_highpage(folio_page(folio, 0));
> + folio_mark_uptodate(folio);
kvm_gmem_mark_prepared() instead of direct folio_mark_uptodate() here, I
think (in preparation of things like [1])? Noticed this while rebasing
my direct map removal series on top of this and wondering why mmap'd
folios sometimes didn't get removed (since it hooks mark_prepared()).
Best,
Patrick
[1]: https://lore.kernel.org/kvm/20241108155056.332412-1-pbonzini@xxxxxxxxxx/
