linux-kernel - [PATCH] [RFC] kernfs: Allow vm_ops->close() if VMA is never split

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:   Sat,  4 Jul 2020 22:56:19 +0200
From:   Richard Weinberger <richard@....at>
To:     linux-kernel@...r.kernel.org
Cc:     tj@...nel.org, gregkh@...uxfoundation.org, ebiederm@...ssion.com,
        dan.j.williams@...el.com, Richard Weinberger <richard@....at>
Subject: [PATCH] [RFC] kernfs: Allow vm_ops->close() if VMA is never split

10 years ago commit a6849fa1f7d7 ("sysfs: Fail bin file mmap if vma close is implemented.")
removed support for vm_ops->close() for mmap on sysfs.
As far I understand the reason is that due to the wrapping in kernfs
every VMA split operation needs to be tracked to call vm_ops->close()
for all fragments. This is not feasible with reasonable effort.

Since commit 31383c6865a5 ("mm, hugetlbfs: introduce ->split() to vm_operations_struct")
we can get notified as soon a VMA is split, this can help to relax the restriction.
So I propose to allow having a custom close under the condition that a
VMA cannot get split.

Signed-off-by: Richard Weinberger <richard@....at>
---
 fs/kernfs/file.c | 42 ++++++++++++++++++++++++++++++++++--------
 1 file changed, 34 insertions(+), 8 deletions(-)

diff --git a/fs/kernfs/file.c b/fs/kernfs/file.c
index 06b342d8462b..82b09e72acff 100644
--- a/fs/kernfs/file.c
+++ b/fs/kernfs/file.c
@@ -347,6 +347,38 @@ static void kernfs_vma_open(struct vm_area_struct *vma)
 	kernfs_put_active(of->kn);
 }
 
+static void kernfs_vma_close(struct vm_area_struct *vma)
+{
+	struct file *file = vma->vm_file;
+	struct kernfs_open_file *of = kernfs_of(file);
+
+	if (!of->vm_ops)
+		return;
+
+	if (!kernfs_get_active(of->kn))
+		return;
+
+	if (of->vm_ops->close)
+		of->vm_ops->close(vma);
+
+	kernfs_put_active(of->kn);
+}
+
+static int kernfs_vma_split(struct vm_area_struct *vma, unsigned long addr)
+{
+	struct file *file = vma->vm_file;
+	struct kernfs_open_file *of = kernfs_of(file);
+
+	/*
+	 * We cannot keep track of split operations,
+	 * so refuse splitting a VMA if someone uses close.
+	 */
+	if (of->vm_ops && of->vm_ops->close)
+		return -EINVAL;
+
+	return 0;
+}
+
 static vm_fault_t kernfs_vma_fault(struct vm_fault *vmf)
 {
 	struct file *file = vmf->vma->vm_file;
@@ -457,6 +489,8 @@ static struct mempolicy *kernfs_vma_get_policy(struct vm_area_struct *vma,
 
 static const struct vm_operations_struct kernfs_vm_ops = {
 	.open		= kernfs_vma_open,
+	.close		= kernfs_vma_close,
+	.split		= kernfs_vma_split,
 	.fault		= kernfs_vma_fault,
 	.page_mkwrite	= kernfs_vma_page_mkwrite,
 	.access		= kernfs_vma_access,
@@ -505,14 +539,6 @@ static int kernfs_fop_mmap(struct file *file, struct vm_area_struct *vma)
 	if (of->mmapped && of->vm_ops != vma->vm_ops)
 		goto out_put;
 
-	/*
-	 * It is not possible to successfully wrap close.
-	 * So error if someone is trying to use close.
-	 */
-	rc = -EINVAL;
-	if (vma->vm_ops && vma->vm_ops->close)
-		goto out_put;
-
 	rc = 0;
 	of->mmapped = true;
 	of->vm_ops = vma->vm_ops;
-- 
2.26.2