KVM: SEV: Pin SEV guest memory out of CMA area

hygon inclusion
category: bugfix
CVE: NA

---------------------------

When pin_user_pages_fast pin SEV guest memory without FOLL_LONGTERM
flag, the pinning pages may be in CMA area, which resulting in other
applications may can't use the CMA area because the pinning pages
can't be migrated.

Add FOLL_LONGTERM flag to pin_user_pages_fast, which makes sure that we
don't keep non_movable pages (due to page reference count) in CMA area.
So CMA area can be allocated by other applications.

Signed-off-by: Xin Jiang <[email protected]>
Signed-off-by: hanliyang <[email protected]>

Author: Xin Jiang

arch/x86/kvm/svm/sev.c

@@ -516,6 +516,7 @@ static struct page **sev_pin_memory(struct kvm *kvm, unsigned long uaddr,
 	unsigned long locked, lock_limit;
 	struct page **pages;
 	unsigned long first, last;
+	unsigned int flags = 0;
 	int ret;
 
 	lockdep_assert_held(&kvm->lock);
@@ -548,8 +549,10 @@ static struct page **sev_pin_memory(struct kvm *kvm, unsigned long uaddr,
 	if (!pages)
 		return ERR_PTR(-ENOMEM);
 
+	flags = write ? FOLL_WRITE : 0;
+
 	/* Pin the user virtual address. */
-	npinned = pin_user_pages_fast(uaddr, npages, write ? FOLL_WRITE : 0, pages);
+	npinned = pin_user_pages_fast(uaddr, npages, flags | FOLL_LONGTERM, pages);
 	if (npinned != npages) {
 		pr_err("SEV: Failure locking %lu pages.\n", npages);
 		ret = -ENOMEM;