[PATCH v1 14/14] mm/mshare: Copy PTEs to host mm

[Khalid Aziz <khalid.aziz@oracle.com>]

VMAs for shared addresses are hosted by a separate host mm. Copy over
original PTEs from the donor process to host mm so the PTEs are
maintained independent of donor process.

Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
---
 include/linux/mm.h |  2 ++
 mm/memory.c        | 48 ++++++++++++++++++++++++++++++++++++++++++++++
 mm/mshare.c        | 14 +++++---------
 3 files changed, 55 insertions(+), 9 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index d9456d424202..78c22891a792 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1845,6 +1845,8 @@ void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
 		unsigned long end, unsigned long floor, unsigned long ceiling);
 int
 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
+int
+mshare_copy_ptes(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
 			  struct mmu_notifier_range *range, pte_t **ptepp,
 			  pmd_t **pmdpp, spinlock_t **ptlp);
diff --git a/mm/memory.c b/mm/memory.c
index e7c5bc6f8836..9010d68f053a 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1234,6 +1234,54 @@ copy_p4d_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
 	return 0;
 }
 
+/*
+ * Copy PTEs for mshare'd pages.
+ * This code is based upon copy_page_range()
+ */
+int
+mshare_copy_ptes(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma)
+{
+	pgd_t *src_pgd, *dst_pgd;
+	unsigned long next;
+	unsigned long addr = src_vma->vm_start;
+	unsigned long end = src_vma->vm_end;
+	struct mm_struct *dst_mm = dst_vma->vm_mm;
+	struct mm_struct *src_mm = src_vma->vm_mm;
+	struct mmu_notifier_range range;
+	int ret = 0;
+
+	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_PAGE,
+				0, src_vma, src_mm, addr, end);
+	mmu_notifier_invalidate_range_start(&range);
+	/*
+	 * Disabling preemption is not needed for the write side, as
+	 * the read side doesn't spin, but goes to the mmap_lock.
+	 *
+	 * Use the raw variant of the seqcount_t write API to avoid
+	 * lockdep complaining about preemptibility.
+	 */
+	mmap_assert_write_locked(src_mm);
+	raw_write_seqcount_begin(&src_mm->write_protect_seq);
+
+	dst_pgd = pgd_offset(dst_mm, addr);
+	src_pgd = pgd_offset(src_mm, addr);
+	do {
+		next = pgd_addr_end(addr, end);
+		if (pgd_none_or_clear_bad(src_pgd))
+			continue;
+		if (unlikely(copy_p4d_range(dst_vma, src_vma, dst_pgd, src_pgd,
+					    addr, next))) {
+			ret = -ENOMEM;
+			break;
+		}
+	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
+
+	raw_write_seqcount_end(&src_mm->write_protect_seq);
+	mmu_notifier_invalidate_range_end(&range);
+
+	return ret;
+}
+
 int
 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma)
 {
diff --git a/mm/mshare.c b/mm/mshare.c
index fba31f3c190f..a399234bf106 100644
--- a/mm/mshare.c
+++ b/mm/mshare.c
@@ -385,7 +385,6 @@ SYSCALL_DEFINE5(mshare, const char __user *, name, unsigned long, addr,
 			 * Copy this vma over to host mm
 			 */
 			vma->vm_private_data = info;
-			vma->vm_mm = new_mm;
 			vma->vm_flags |= VM_SHARED_PT;
 			new_vma = vm_area_dup(vma);
 			if (!new_vma) {
@@ -394,6 +393,7 @@ SYSCALL_DEFINE5(mshare, const char __user *, name, unsigned long, addr,
 				err = -ENOMEM;
 				goto free_info;
 			}
+			new_vma->vm_mm = new_mm;
 			err = insert_vm_struct(new_mm, new_vma);
 			if (err) {
 				mmap_write_unlock(new_mm);
@@ -402,17 +402,13 @@ SYSCALL_DEFINE5(mshare, const char __user *, name, unsigned long, addr,
 				goto free_info;
 			}
 
+			/* Copy over current PTEs */
+			err = mshare_copy_ptes(new_vma, vma);
+			if (err != 0)
+				goto free_info;
 			vma = vma->vm_next;
 		}
 
-		/*
-		 * Copy over current PTEs
-		 */
-		myaddr = addr;
-		while (myaddr < new_mm->task_size) {
-			*pgd_offset(new_mm, myaddr) = *pgd_offset(old_mm, myaddr);
-			myaddr += PGDIR_SIZE;
-		}
 		/*
 		 * TODO: Free the corresponding page table in calling
 		 * process
-- 
2.32.0