Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
In-Reply-To: <CAOQ4uxg6hVd2k5T20N1u72Of9Eb3O_EAasC0MhDoLY3-o2UrGw@mail.gmail.com>
References: <20180529144143.16378-1-mszeredi@redhat.com> <20180529144143.16378-8-mszeredi@redhat.com>
 <CAOQ4uxg6hVd2k5T20N1u72Of9Eb3O_EAasC0MhDoLY3-o2UrGw@mail.gmail.com>
From:   Miklos Szeredi <miklos@szeredi.hu>
Date:   Wed, 30 May 2018 10:18:03 +0200
Message-ID: <CAJfpegv1HLghFSczd1oZbrpOaz39r8VD5M47aH1eVJFQxs=keg@mail.gmail.com>
Subject: Re: [PATCH 07/11] ovl: make ovl_create_real() cope with vfs_mkdir() safely
To:     Amir Goldstein <amir73il@gmail.com>
Cc:     Miklos Szeredi <mszeredi@redhat.com>,
        overlayfs <linux-unionfs@vger.kernel.org>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, May 29, 2018 at 5:29 PM, Amir Goldstein <amir73il@gmail.com> wrote:
> On Tue, May 29, 2018 at 5:41 PM, Miklos Szeredi <mszeredi@redhat.com> wrote:
>> From: Amir Goldstein <amir73il@gmail.com>
>>
>> vfs_mkdir() may succeed and leave the dentry passed to it unhashed and
>> negative.  ovl_create_real() is the last caller breaking when that
>> happens.
>>
>> [amir: split re-factoring of ovl_create_temp() to prep patch
>>        add comment about unhashed dir after mkdir
>>        add pr_warn() if mkdir succeeds and lookup fails]
>>
>> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
>> Signed-off-by: Amir Goldstein <amir73il@gmail.com>
>> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
>> ---
>>  fs/overlayfs/dir.c | 35 +++++++++++++++++++++++++++++++++--
>>  1 file changed, 33 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/overlayfs/dir.c b/fs/overlayfs/dir.c
>> index 1b181292a624..b33d37d1a87c 100644
>> --- a/fs/overlayfs/dir.c
>> +++ b/fs/overlayfs/dir.c
>> @@ -114,6 +114,37 @@ int ovl_cleanup_and_whiteout(struct dentry *workdir, struct inode *dir,
>>         goto out;
>>  }
>>
>> +static struct dentry *ovl_mkdir_real(struct inode *dir, struct dentry *dentry,
>> +                                    umode_t mode)
>> +{
>> +       int err;
>> +
>> +       err = ovl_do_mkdir(dir, dentry, mode);
>> +       if (err) {
>> +               dput(dentry);
>> +               return ERR_PTR(err);
>> +       }
>> +
>> +       /*
>> +        * vfs_mkdir() may succeed and leave the dentry passed
>> +        * to it unhashed and negative. If that happens, try to
>> +        * lookup a new hashed and positive dentry.
>> +        */
>> +       if (unlikely(d_unhashed(dentry))) {
>> +               struct dentry *d;
>> +
>> +               d = lookup_one_len(dentry->d_name.name, dentry->d_parent,
>> +                                  dentry->d_name.len);
>> +               if (IS_ERR(d)) {
>> +                       pr_warn("overlayfs: failed lookup after mkdir (%pd2, err=%i).\n",
>> +                               dentry, err);
>> +               }
>> +               dput(dentry);
>> +               dentry = d;
>> +       }
>> +       return dentry;
>> +}
>> +
>>  struct dentry *ovl_create_real(struct inode *dir, struct dentry *newdentry,
>>                                struct ovl_cattr *attr)
>>  {
>> @@ -135,8 +166,8 @@ struct dentry *ovl_create_real(struct inode *dir, struct dentry *newdentry,
>>                         break;
>>
>>                 case S_IFDIR:
>> -                       err = ovl_do_mkdir(dir, newdentry, attr->mode);
>> -                       break;
>> +                       /* mkdir is special... */
>> +                       return ovl_mkdir_real(dir, newdentry, attr->mode);
>>
>
> I like your version with the helper.
> So do we not care about the WARN_ON below on non-instantiated dentry
> for directories? or we don't need it at all?

I think we do.  Filesystems do all sorts of weird an unexpected
things.   We should definitely not assume underlying filesystem
returns in any way sane result.

Fixed, thanks for spotting.

Miklos