| Message ID | 20211119134739.20218-5-chao.p.peng@linux.intel.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | KVM: mm: fd-based approach for supporting KVM guest private memory | expand |
On 11/19/21 14:47, Chao Peng wrote: > For fd-based memslot store the file references for shared fd and the > private fd (if any) in the memslot structure. Since there is no 'hva' > concept we cannot call hva_to_pfn() to get a pfn, instead kvm_memfd_ops > is added to get_pfn/put_pfn from the memory backing stores that provide > these fds. > > Signed-off-by: Yu Zhang<yu.c.zhang@linux.intel.com> > Signed-off-by: Chao Peng<chao.p.peng@linux.intel.com> > --- What about kvm_read/write_guest? Maybe the proposal which kept userspace_addr for the shared fd is more doable (it would be great to ultimately remove the mandatory userspace mapping for the shared fd, but I think KVM is not quite ready for that). Paolo
On Tue, Nov 23, 2021 at 09:41:34AM +0100, Paolo Bonzini wrote: > On 11/19/21 14:47, Chao Peng wrote: > > For fd-based memslot store the file references for shared fd and the > > private fd (if any) in the memslot structure. Since there is no 'hva' > > concept we cannot call hva_to_pfn() to get a pfn, instead kvm_memfd_ops > > is added to get_pfn/put_pfn from the memory backing stores that provide > > these fds. > > > > Signed-off-by: Yu Zhang<yu.c.zhang@linux.intel.com> > > Signed-off-by: Chao Peng<chao.p.peng@linux.intel.com> > > --- > > What about kvm_read/write_guest? Hmm, that would be another area KVM needs to change. Not totally undoable. > Maybe the proposal which kept > userspace_addr for the shared fd is more doable (it would be great to > ultimately remove the mandatory userspace mapping for the shared fd, but I > think KVM is not quite ready for that). Agree for short term keeping shared part unchanged would be making work easy:) Let me try that to see if any blocker. > > Paolo
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h index 99e9f9969703..1d4ac0c9b63b 100644 --- a/include/linux/kvm_host.h +++ b/include/linux/kvm_host.h @@ -424,6 +424,12 @@ static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu) */ #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1) +struct kvm_memfd_ops { + kvm_pfn_t (*get_pfn)(struct kvm_memory_slot *slot, struct file *file, + gfn_t gfn, bool alloc, int *order); + void (*put_pfn)(kvm_pfn_t pfn); +}; + struct kvm_memory_slot { gfn_t base_gfn; unsigned long npages; @@ -433,6 +439,9 @@ struct kvm_memory_slot { u32 flags; short id; u16 as_id; + struct file *file; + struct file *priv_file; + struct kvm_memfd_ops *memfd_ops; }; static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot) @@ -1310,6 +1319,20 @@ static inline int memslot_id(struct kvm *kvm, gfn_t gfn) return gfn_to_memslot(kvm, gfn)->id; } +static inline bool memslot_is_memfd(const struct kvm_memory_slot *slot) +{ + if (slot && slot->memfd_ops) + return true; + return false; +} + +static inline bool memslot_has_private(const struct kvm_memory_slot *slot) +{ + if (slot && slot->priv_file) + return true; + return false; +} + static inline gfn_t hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot) {