From: Matthew Wilcox (Oracle) <willy@infradead.org>
Date: Wed, 30 Sep 2020 16:49:02 +0000 (-0400)
Subject: start of reparenting VMAs.  buggy.
X-Git-Url: https://www.infradead.org/git/?a=commitdiff_plain;h=refs%2Fheads%2Fmshare;p=users%2Fwilly%2Flinux.git

start of reparenting VMAs.  buggy.
---

diff --git a/mm/mmap.c b/mm/mmap.c
index 8c7ca737a19b..376837e90b4b 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -482,8 +482,7 @@ static __always_inline void vma_rb_erase_ignore(struct vm_area_struct *vma,
 	__vma_rb_erase(vma, root);
 }
 
-static __always_inline void vma_rb_erase(struct vm_area_struct *vma,
-					 struct rb_root *root)
+void vma_rb_erase(struct vm_area_struct *vma, struct rb_root *root)
 {
 	/*
 	 * All rb_subtree_gap values must be consistent prior to erase,
diff --git a/mm/mshare.c b/mm/mshare.c
index f79e6519c94a..618912b1d14e 100644
--- a/mm/mshare.c
+++ b/mm/mshare.c
@@ -71,13 +71,65 @@ static const struct file_operations mshare_fops = {
 	.release = mshare_release,
 };
 
+void vma_rb_erase(struct vm_area_struct *vma, struct rb_root *root);
+
+static
+bool move_vmas(struct mm_struct *new_mm, unsigned long addr, unsigned long len)
+{
+	struct mm_struct *old_mm = current->mm;
+	struct vm_area_struct *vma, *next, *prev = NULL;
+	bool success = false;
+	int i = 0;
+
+	mmap_write_lock(old_mm);
+
+	/* Must not overlap end */
+	vma = find_vma(old_mm, addr + len);
+	if (vma && vma->vm_start < addr + len)
+		goto unlock;
+	vma = find_vma(old_mm, addr);
+	if (vma && vma->vm_start < addr)
+		goto unlock;
+	success = true;
+
+	if (vma)
+		prev = vma->vm_prev;
+	while (vma && vma->vm_start < addr + len) {
+		int ret;
+
+printk("found vma start %lx\n", vma->vm_start);
+		vma_rb_erase(vma, &old_mm->mm_rb);
+		old_mm->map_count--;
+
+		next = vma->vm_next;
+		ret = insert_vm_struct(new_mm, vma);
+		if (ret)
+			goto unlock;
+		/* XXX: VM_ACCOUNT? */
+		vm_stat_account(new_mm, vma->vm_flags, -vma_pages(vma));
+		vma = next;
+	}
+	if (prev)
+		prev->vm_next = vma;
+	if (vma)
+		vma->vm_prev = prev;
+
+	while (addr < new_mm->task_size) {
+		new_mm->pgd[i++] = *pgd_offset(old_mm, addr);
+		addr += PGDIR_SIZE;
+	}
+unlock:
+	mmap_write_unlock(old_mm);
+
+	return success;
+}
+
 SYSCALL_DEFINE3(mshare, unsigned long, addr, unsigned long, len,
 		unsigned long, flags)
 {
 	struct mm_struct *mm;
-	struct vm_area_struct *vma;
+	struct file *file;
 	int fd;
-	int i = 0;
 
 	if ((addr | len) & (PGDIR_SIZE - 1))
 		return -EINVAL;
@@ -92,30 +144,24 @@ SYSCALL_DEFINE3(mshare, unsigned long, addr, unsigned long, len,
 	if (!mm->task_size)
 		mm->task_size--;
 
-	mmap_write_lock(current->mm);
-
-	vma = find_vma(current->mm, addr + len);
-	if (vma && vma->vm_start < addr + len)
-		goto unlock;
-	vma = find_vma(current->mm, addr);
-	if (vma && vma->vm_start < addr)
-		goto unlock;
-
-	while (addr < mm->task_size) {
-		mm->pgd[i++] = *pgd_offset(current->mm, addr);
-		addr += PGDIR_SIZE;
-	}
-	mmap_write_unlock(current->mm);
-
-	fd = anon_inode_getfd("mshare", &mshare_fops, mm, O_RDWR);
+	fd = get_unused_fd_flags(O_RDWR);
 	if (fd < 0)
-		goto nofd;
+		goto out;
+	file = anon_inode_getfile("mshare", &mshare_fops, mm, O_RDWR);
+	if (IS_ERR(file))
+		goto nofile;
+
+	if (!move_vmas(mm, addr, len))
+		goto invalid;
+	fd_install(fd, file);
 out:
 	return fd;
-unlock:
-	mmap_write_unlock(current->mm);
-	fd = -EINVAL;
-nofd:
+invalid:
+	fput(file);
+	file = ERR_PTR(-EINVAL);
+nofile:
+	put_unused_fd(fd);
+	fd = PTR_ERR(file);
 	mmput(mm);
 	goto out;
 }