linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Linus Torvalds <torvalds@linux-foundation.org>
To: Al Viro <viro@zeniv.linux.org.uk>
Cc: syzbot <syzbot+7a8ba368b47fdefca61e@syzkaller.appspotmail.com>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	Linux List Kernel Mailing <linux-kernel@vger.kernel.org>,
	syzkaller-bugs <syzkaller-bugs@googlegroups.com>
Subject: Re: KASAN: use-after-free Read in path_lookupat
Date: Mon, 25 Mar 2019 12:18:02 -0700	[thread overview]
Message-ID: <CAHk-=whJ4M5FegOLvnjUtJ0+pHv4L8UNFt+9jJDhox3Ada8kwA@mail.gmail.com> (raw)
In-Reply-To: <CAHk-=wg4iJsMHBzK52WzP+5_92HbwvX_vh_s4mMUuN0FJGdM5A@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 678 bytes --]

On Mon, Mar 25, 2019 at 11:36 AM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
> >
> > You mean, split ->destroy_inode() into immediate and RCU-delayed parts?
> > There are filesystems where both parts are non-empty - we can't just
> > switch all ->destroy_inode() work to call_rcu().
>
> Right. Not just move the existing destroy_inode() - because as you
> say, people may not be able to to do that in RCU contect, but split it
> up, and add a "final_free_inode()" callback or something for the RCU
> phase.

Something like the attached.

COMPLETELY UNTESTED. And no filesystems converted to actually use the
new rcu_destroy_inode() thing.

Hmm?

                Linus

[-- Attachment #2: patch.diff --]
[-- Type: text/x-patch, Size: 3424 bytes --]

 Documentation/filesystems/vfs.txt |  6 ++++++
 fs/inode.c                        | 27 +++++++++++++++++++++------
 include/linux/fs.h                |  1 +
 3 files changed, 28 insertions(+), 6 deletions(-)

diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index 761c6fd24a53..60f7841a12e6 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -210,6 +210,7 @@ filesystem. As of kernel 2.6.22, the following members are defined:
 struct super_operations {
         struct inode *(*alloc_inode)(struct super_block *sb);
         void (*destroy_inode)(struct inode *);
+        int (*rcu_destroy_inode)(struct inode *);
 
         void (*dirty_inode) (struct inode *, int flags);
         int (*write_inode) (struct inode *, int);
@@ -248,6 +249,11 @@ or bottom half).
   	->alloc_inode was defined and simply undoes anything done by
 	->alloc_inode.
 
+ rcu_destroy_inode: this method is called after the RCU delay by
+	destroy_inode() to release resources allocated for struct inode.
+	If it returns a non-zero value, it means that it has free'd the
+	inode, otherwise the inode layer will free it.
+
   dirty_inode: this method is called by the VFS to mark an inode dirty.
 
   write_inode: this method is called when the VFS needs to write an
diff --git a/fs/inode.c b/fs/inode.c
index e9d97add2b36..b85457baad20 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -215,9 +215,13 @@ static struct inode *alloc_inode(struct super_block *sb)
 		return NULL;
 
 	if (unlikely(inode_init_always(sb, inode))) {
-		if (inode->i_sb->s_op->destroy_inode)
+		if (inode->i_sb->s_op->destroy_inode) {
 			inode->i_sb->s_op->destroy_inode(inode);
-		else
+			if (!inode->i_sb->s_op->rcu_destroy_inode)
+				return NULL;
+		}
+		if (!inode->i_sb->s_op->rcu_destroy_inode ||
+		    !inode->i_sb->s_op->rcu_destroy_inode(inode))
 			kmem_cache_free(inode_cachep, inode);
 		return NULL;
 	}
@@ -256,17 +260,28 @@ EXPORT_SYMBOL(__destroy_inode);
 static void i_callback(struct rcu_head *head)
 {
 	struct inode *inode = container_of(head, struct inode, i_rcu);
-	kmem_cache_free(inode_cachep, inode);
+
+	if (!inode->i_sb->s_op->rcu_destroy_inode ||
+	    !inode->i_sb->s_op->rcu_destroy_inode(inode))
+		kmem_cache_free(inode_cachep, inode);
 }
 
 static void destroy_inode(struct inode *inode)
 {
 	BUG_ON(!list_empty(&inode->i_lru));
 	__destroy_inode(inode);
-	if (inode->i_sb->s_op->destroy_inode)
+
+	/*
+	 * If we have a 'destroy_inode' but no 'rcu_destroy_inode'
+	 * then the filesystem handles the RCU-delayed destruction
+	 * on its own, and we don't do any RCU callbacks.
+	 */
+	if (inode->i_sb->s_op->destroy_inode) {
 		inode->i_sb->s_op->destroy_inode(inode);
-	else
-		call_rcu(&inode->i_rcu, i_callback);
+		if (!inode->i_sb->s_op->rcu_destroy_inode)
+			return;
+	}
+	call_rcu(&inode->i_rcu, i_callback);
 }
 
 /**
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 8b42df09b04c..727561ecbc23 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1900,6 +1900,7 @@ extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
 struct super_operations {
    	struct inode *(*alloc_inode)(struct super_block *sb);
 	void (*destroy_inode)(struct inode *);
+	int (*rcu_destroy_inode)(struct inode *);
 
    	void (*dirty_inode) (struct inode *, int flags);
 	int (*write_inode) (struct inode *, struct writeback_control *wbc);

  reply	other threads:[~2019-03-25 19:18 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-11-28 17:40 KASAN: use-after-free Read in path_lookupat syzbot
2019-03-25  0:44 ` syzbot
2019-03-25  1:25   ` Linus Torvalds
2019-03-25  1:23 ` Linus Torvalds
2019-03-25  4:57   ` Al Viro
2019-03-25  9:15     ` Daniel Borkmann
2019-03-25 11:11       ` Al Viro
2019-03-25 11:17         ` Al Viro
2019-03-25 11:21           ` Daniel Borkmann
2019-03-25 18:36     ` Linus Torvalds
2019-03-25 19:18       ` Linus Torvalds [this message]
2019-03-25 21:14         ` Al Viro
2019-03-25 21:45           ` Linus Torvalds
2019-03-25 22:04             ` Daniel Borkmann
2019-03-25 22:13               ` Linus Torvalds
2019-03-25 22:41                 ` Daniel Borkmann
2019-03-25 22:49               ` Al Viro
2019-03-25 23:37             ` Al Viro
2019-03-25 23:44               ` Alexei Starovoitov
2019-03-26  0:21                 ` Al Viro
2019-03-26  1:38               ` ceph: fix use-after-free on symlink traversal Al Viro
2019-03-26  1:39                 ` jffs2: " Al Viro
2019-03-26  1:40                 ` ubifs: " Al Viro
2019-03-26  1:43                 ` debugfs: " Al Viro
2019-03-26 10:41                 ` ceph: " Jeff Layton
2019-03-26 11:38                 ` Ilya Dryomov
2019-03-26  1:45               ` KASAN: use-after-free Read in path_lookupat Al Viro
2019-04-10 18:11                 ` Al Viro
2019-04-10 19:44                   ` Linus Torvalds
2019-03-25 19:43       ` Al Viro
2019-03-25 22:48         ` Dave Chinner
2019-03-25 23:02           ` Al Viro
     [not found]             ` <CAGe7X7mb=gK7zhSwmT_6mmmkcbjhZAOb=wj31BdUcHkNUPsm2Q@mail.gmail.com>
2019-03-26  4:15               ` Al Viro
2019-03-27 16:58                 ` Jan Kara
2019-03-27 18:59                   ` Al Viro
2019-03-28  9:00                     ` Jan Kara
2019-03-27 17:22             ` Jan Kara

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAHk-=whJ4M5FegOLvnjUtJ0+pHv4L8UNFt+9jJDhox3Ada8kwA@mail.gmail.com' \
    --to=torvalds@linux-foundation.org \
    --cc=ast@kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=syzbot+7a8ba368b47fdefca61e@syzkaller.appspotmail.com \
    --cc=syzkaller-bugs@googlegroups.com \
    --cc=viro@zeniv.linux.org.uk \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).