linux-kernel - Re: [rfc][possible solution] RCU vfsmounts

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

Date:	Sun, 29 Sep 2013 19:49:18 +0100
From:	Al Viro <viro@...IV.linux.org.uk>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	linux-fsdevel <linux-fsdevel@...r.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Miklos Szeredi <miklos@...redi.hu>
Subject: Re: [rfc][possible solution] RCU vfsmounts

On Sun, Sep 29, 2013 at 07:10:47PM +0100, Al Viro wrote:
> FWIW, right now I'm reviewing the subset of fs code that can be hit in
> RCU mode.  Not a pretty sight, that... ;-/  First catch: in
> fuse_dentry_revalidate() we have a case (reachable with LOOKUP_RCU) where
> we do this:
>         } else if (inode) {
>                 fc = get_fuse_conn(inode);
>                 if (fc->readdirplus_auto) {
>                         parent = dget_parent(entry);
>                         fuse_advise_use_readdirplus(parent->d_inode);
>                         dput(parent);
>                 }
>         }
> First of all, that'll lead to obvious nastiness if we get here when
> ->s_fs_info has already been freed in process of fs shutdown; fc will
> be pointing to kfreed object and no, freeing it isn't RCU-delayed.
> That's not a problem with the current tree, of course, but this
> dput(parent) very much is - doing that under rcu_read_lock() is
> a Bloody Bad Idea(tm).

Another one: 
int ll_revalidate_nd(struct dentry *dentry, unsigned int flags)
{
        struct inode *parent = dentry->d_parent->d_inode;
        int unplug = 0;

        CDEBUG(D_VFSTRACE, "VFS Op:name=%s,flags=%u\n",  
               dentry->d_name.name, flags);

        if (!(flags & (LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE)) &&
            ll_need_statahead(parent, dentry) > 0) {
                if (flags & LOOKUP_RCU)
                        return -ECHILD;

... and ll_need_statahead(NULL, ...) will oops.  Doesn't even need
LOOKUP_RCU to barf - ->d_revalidate() can be called without ->i_mutex
on parent, so we can race with e.g. rename() followed by rmdir() of
old parent.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/